↳ THE DOCUMENTATION
Develop the idea.
Keep the reasoning.
A practical guide to agentic brainstorming: collect scattered thoughts, investigate with agents, validate assumptions and structure the findings for a builder.
Every chapter names the exact buttons, panels and fields you will see, and shows them in use on a running example: could cooler streets support local cafés? The same example ships inside the workspace, so you can follow along on a real canvas.
One canvas.
An agent beside you.
Synatrail is a canvas workspace for agentic brainstorming, ideation and specification building. Start with a thought, an image, a question or a piece of data. Connect the fragments, ask agents to research and challenge them, and review what the evidence supports. Return to earlier questions as the idea develops. When the scope is clear, prepare a specification that carries your decisions and acceptance criteria into the build.
The guided tour
On your first visit, the guided tour points out the project, cards, connections, model access and notebook. You can close it and restart it from Help (?). The working canvas stays available while you learn.
Your canvas and the agent's map
The investigation opens on one working canvas. Add your question and evidence there. Once the agent is watching, its developing hypotheses, directions, tests and results also become visible on that canvas, with their agent origin and tentative status shown on each card. The agent keeps its own map in the background; you decide what to explore, ignore, edit or accept as your own judgment. A computed verdict does not set your verdict.
The Agent button opens the rail for discoveries, activity, decisions, automation and the notebook. The presence control above the canvas reads Following this investigation while the watch is active. The assistant at the bottom has separate Ask and Steer modes.
Open any figure for a full-size view. On a phone, wide desktop captures scroll sideways; the first figure switches to the real phone layout.

Seven words to know
- Investigation, also called a project: one question, one working canvas, and its evidence, agent activity and history.
- Node: a card on the canvas. A paper, a dataset, a question, an idea, a parameter, an experiment.
- Relationship: a typed edge between two nodes. The interface calls it a hypothesis, because it is one.
- Agent watch: the standing instruction to follow this investigation within the automation level and daily budget you set.
- Agent idea: a hypothesis, direction, test or candidate developed by the agent and marked tentative until evidence and your review settle it.
- Publication: a frozen, reviewed copy of My canvas, optionally with notebook entries, visible to signed-in users.
- Produce package: a versioned specification handoff from selected evidence, with artifact checks, archives and an exact release manifest. Chapter 15. Earlier planning packages remain readable.
Chapter 17 has the full glossary.
Getting in
The workspace lives at synatrail.com/app/. Registration is open: anyone who completes the form gets an account at once, with no approval step. Your investigations are private until you choose what to share, and the Terms of Use and Privacy Notice say exactly what is stored and why.
- Create your account
Use Create an account on the sign-in page. The form is titled Create your account and asks for your Email, a Password of at least twelve characters and Your name; then, optionally, Organisation, Role and What do you want to investigate?, whose placeholder reads A question, a domain, or what brought you here. Tick I accept the Terms of Use and acknowledge the Privacy Notice and finish with Create account →. Your workspace opens right away.
- Optional privacy choices
Your account's required terms acceptance is separate from optional product news and pseudonymised research choices. Leave those optional boxes unticked if you do not want them; you can change them later in Account → Privacy choices. Browser analytics has its own Privacy settings.
- Sign in
The sign-in form is titled Sign in to Synatrail. Enter your email and password and press Sign in →. Forgot your password? Ask your administrator for a reset link; it works once, for 30 minutes.
- Open the example
Inside, press Current project, expand More in this project and choose Open an illustrative example. The project is called Illustrative example · Cooler streets. Every node in it is marked Illustrative · unverified, and every edge says Illustrative connection. Replace with a cited rationale. This documentation tells the same story with real steps.
Main actions have the displayed credit fees described in chapter 17. With BYOM, your model provider bills usage through your key or plan. When qualified Synatrail-managed intelligence is available, its model usage is charged in the same platform credits at the displayed price based on actual tokens. Nothing reaches a model until you choose a source and consent in chapter 07. Browsing and working on your board remain free.
Give the question
a place to live.
A project is one investigation: its working canvas, sources, notebook, experiments and research history. The header holds everything you need to manage projects.
The header, left to right
- Current project shows the open project's title. Press it to open its details.
- Save (⌘ or Ctrl + S), New and Projects, which opens your investigation library.
- The save indicator: ✓ Saved, or ○ Saving, Unsaved, Conflict, Needs attention or Offline. Chapter 12 explains each state.
- The package action is Produce package when the comprehensive workbench is enabled, or Build package for the earlier composer (chapter 15). On the right are Public investigations, Share, the theme toggle and your avatar.

Begin with your intent
New opens What is this investigation for?. Write the intent or direction in one or two sentences and optionally Add a file. Choose a Package direction (optional): None, Scientific, Software, Hardware, Hybrid, Financial or Services. None keeps research open without a package objective. The other choices guide the agent toward the evidence and deliverables for that package. Your choice stays with the investigation; it does not create a package or change the automation policy. Open the canvas myself creates a manual investigation. Continue with the agent saves a Full auto policy with the initial daily budget shown in the Agent controls; missing model access opens settings. Review its saved capabilities and budget before relying on background work. New users can also choose a ready-made starter template.
Project details
Current project opens the panel Project details. It holds the two things a project needs: Project name and The question you are following. Press Save project to store them. Changing an existing question substantively shows Why the reframe? (optional); saving then records a Reframe entry in the notebook with the former and new wording, and the notebook entry opens project details. Canvas changes save on their own, and the strip at the top says so: Canvas changes save automatically. Under FROM INVESTIGATION TO BUILD sit Make a build package and Idea library, both explained in chapter 15. The More in this project disclosure links to All investigations, folders & Trash, Branches & comparisons, Sources & experiments, Community and Open an illustrative example.
The investigation library
Projects in the header opens a full-screen library titled Investigations. Its sidebar has All investigations, Unfiled, Idea library, a FOLDERS list with a New folder button, and Trash, each with a count. Above the list: New investigation, a search field, Search names or questions, a sort, Recently updated or Name A–Z, and a refresh button. Each card shows a drawn preview of the canvas, the title, a ✓ Current badge on the open project, its question, its folder, the node count and the last change, with Open, Rename, Move (to a folder or back to Unfiled) and Delete. Deleting asks Move to Trash? and warns that active research will stop and the public snapshot will be withdrawn; Restore in Trash brings the investigation back, still private and with research stopped. Folders offer Rename folder and Delete folder, which moves their investigations to Unfiled. Esc or Back to canvas closes the library.
Switching, naming, exporting
New projects are named Untitled project, then Untitled project 2, and so on, until you rename them. You cannot switch projects while edits are waiting to be saved. The message reads Save or resolve this investigation's pending edits before switching.
Export canvas in the status bar downloads the investigation canvas and your notes as a JSON file. Export workspace, in Settings under Account, exports everything you own. These files are records for you. There is no importer for them. Evidence enters the workspace through uploads, links and data connections.
Start with a piece
of something.
A node is one card on the canvas. Anything you know, suspect or want to test becomes a node. There are four ways to add one, and one panel that shapes it.
Four ways to add
- Press the + tool in the floating rail, whose full name is Add source or node (+), or press + on the keyboard.
- Double-click an empty spot on the canvas.
- Drag from a node's round output on the right and release on empty canvas. A new node appears already connected.
- Drop files anywhere, or paste. A pasted web address becomes a source node, pasted text becomes a context note, and pasted files upload.
The first three open the Start with a piece of something. menu:
| Entry | What the menu says | What it creates |
|---|---|---|
| Drop a document | PDF, text, spreadsheet or image · up to 25 MB | A file picker. The upload becomes a source node while extraction runs. |
| Connect an API | A data source with its own connection | A Data connection node and the connector panel of chapter 05. |
| Set a starting point | A doubt, a curiosity or a thread to pursue | A Starting point node titled What are you curious about? |
| Add context | An observation, constraint or rough idea | A Context node titled A piece of context. |
| Add a dataset | Describe a dataset or setup manifest and connect it | A Dataset card, ready to name and edit. |
| Record an observation | Keep a finding, event or staff record | An Observation card for your own notes. |
| Add an interpretation | Explain what the evidence means or what needs review | An Interpretation card for your reading of the evidence. |
| Bring a link | A page, paper or source to investigate | A Source node titled A source worth exploring. |
| Try a parameter | Change an input and compare what happens | A Change an input node titled Change an assumption, value 28 in a 0 to 100 range. |
| Capture an idea | A possibility worth testing | A Possible direction node titled A possible direction. |
| Create an experiment | Run a defined operation on your evidence | An Experiment node titled A test worth running. Chapter 11. |
| Group evidence | Name a shared question and the sources behind it | A Research group frame titled A research frame. Chapter 06. |
On ordinary cards, Explore opens the assistant for that card. The separate arrow button, Choose next step, opens the next-step choices.
The menu ends with a link to Source catalog & saved work, the panel called Bring in what you already know., where the field Find saved evidence searches sources, notes and tests already in this project and marks those that are on canvas.
Every kind of node
The icon, outline and label identify a card’s role: folded documents for sources, a cylinder for datasets, a compact control for parameters, a flask for experiments and a chart for results. Some kinds are created by you, some by the agent, some by experiments.
Start describes the beginning of a workflow. A supporting input or association does not turn it into a continuation. Its ordinary left flow input stays hidden until you explicitly review Join as continuation; named supporting inputs remain available. Older canvases keep their original relationship rules until you choose Examine conversion and mark which relationships are workflow paths.
| Card label | Used for | Usually created by |
|---|---|---|
| Starting point | The doubt, curiosity or thread that begins the investigation, or a sub-question along the way. | You |
| Paperalso Study · Patent | Published work, with attribution and rights notes. | You, or the agent when it proposes a reading |
| Source | A web page or other public source to read. | You, by link, or the agent |
| Dataset | A CSV or JSON upload with immutable versions. | Uploads |
| Data connection | A bounded API read with its own credentials. | You |
| Contextalso Observation | An observation, a constraint, a rough note. | You, or pasted text |
| Interpretationalso Claim | A reading of the evidence, stated as a claim. | You, or the agent |
| Possible direction | An idea worth a test. | You, or the agent |
| Answer | A saved assistant answer with its evidence references. | Saved from a conversation, chapter 08 |
| Change an input | A parameter: a value, a unit and a range you can slide. | You |
| Experiment | One of the nine numerical tools with pinned inputs. | You, chapter 11 |
| Computed result | The output of a run, with its calculation receipt. | The computation worker |
| Research group | A frame around related nodes with a shared purpose. | You, chapter 06 |
Reasoning roles and supply cards
Configured reasoning steps show Clarify, Explore, Challenge, Design a test, Compare or Synthesize instead of their underlying card kind. Other typed roles include Material collection, Transform, Route, AI investigator, Experiment and Product specification. Supply cards provide Evidence, Data, Model, Tools, Memory, Settings or Criteria through named inputs. A visible role can be readable while its executor is unavailable; the inspector explains the difference.

Show the current card editor

Shape a node
Press Enter on a focused card, double-click it, or use its Open button to open the node panel, titled A piece of the investigation. Its fields:
- Title, up to 200 characters. On a question card this field is called Starting point, and the panel is titled Where the investigation begins.
- A body. For questions it is called Context: the place for what you already know, what puzzles or frustrates you, where you feel lost, and what you hope to find out. For everything else, What should the agent understand? Up to 18,000 characters.
- The optional Capture context disclosure on starting points, notes and source-like cards: What prompted this? keeps the moment or stimulus and Why this caught my attention keeps your personal reason. Both stay separate from source text, summaries and extracted facts.
- For parameters, a grid: Value, Unit (the placeholder suggests °C, EUR, kg or 1), Minimum, Maximum and Step.
- Source link, a web address.
- The Attribution & assessment disclosure: Tags, Who, where and when / source on capture cards (or Source / author elsewhere), Rights notes and Evidence status.
Save on canvas stores the card, with the toast Node saved. On a starting point, Capture what prompted this adds a context card and a raises question connection in one undoable edit. On an existing capture card, This prompted my question adds that relationship only when you choose it; ordinary paste, drop and add gestures leave the card unconnected. Explore this sends the node to the assistant with the question What is worth investigating about this node? Propose the sub-questions, the evidence to gather and the next steps as cards. The trash button, Delete node, removes it. Undo brings it back. A locked card first asks you to unlock it.
Parameter cards show their value and unit under the word Assumption, with a slider you can drag on the card itself. Source cards add Open original source, View uploaded original and the extraction tools of chapter 05.
What connects
these two?
Connections are where the workspace earns its keep. Every edge has a type, a rationale and an assessment. None of them is proof. The edge panel says so in its footer: A connection is not proof of causation.
Three ways to connect
- Drag. Drag from a compatible output to an input. Ports show the role and kind of material they accept. A Start has no ordinary left flow input; joining it as a continuation requires an explicit review. Named supporting inputs stay available.
- Keyboard. Focus a port and press Enter or Space. The toast says Choose another node's handle to connect. Escape cancels. Then activate the other card's port.
- The Connect tool. Press C or the rail button Connect nodes (C), then choose an output and an input. With exactly two cards selected, the selection toolbar also offers Connect.
Three connection channels
Forward flow orders reasoning steps. Supporting input supplies evidence, data, settings, criteria, context, model or tools to a named socket. Association records a relationship. An association alone does not schedule work, and a model or tool attachment does not grant permission to use it.
The card's Connection points section lists required, missing, connected, full and unavailable sockets. Connect only compatible material. If a connection is rejected, the panel explains the reason and keeps your first socket selected. A Route has explicit Continue and Revisit outcomes; a product specification needs a separately configured handoff before it has an onward output.
Connections keep their exact sockets and channel. Inspect the connection you want to change rather than assuming every pair of cards has only one relationship.

Show the current relationship editor

The edge panel
Its overline reads A RELATIONSHIP IS A HYPOTHESIS and its title What connects these? Eight quick chips sit at the top: related, informs, supports, contradicts, analogy, depends on, feeds input and joins on. The Relationship dropdown holds all twenty-three types.
| Family | Types | Use it when |
|---|---|---|
| Evidence | supports · contradicts · qualifies · informs · related | One node bears on another's credibility, or simply belongs nearby. |
| Reasoning | raises question · could explain · predicts · similar mechanism · different application · analogy | You are stating how one thing might account for, or resemble, another. |
| Structure | depends on · requires · enables · limits · located within · aligned in time | Constraints, prerequisites and shared place or time. |
| Data | joins on · candidate join · feeds input · transforms · aggregates · evaluates | Datasets, parameters and experiments passing values to each other. |
Use Text on the connection to add up to 240 characters below a relationship’s type on the canvas. Save it with Save relationship. Clear the field to remove the annotation; the relationship type and its rationale stay separate.
Below the type, the field Why do you think they connect? takes up to 2,000 characters of rationale. Assessment is one of unassessed, tentative, supported or refuted. Reverse direction swaps the arrow. Save relationship stores it.
Interpret asks the assistant about the edge. Challenge and Design a test configure reasoning steps against its exact revision. Make testable opens your own Experiment plan. None of these actions changes the assessment automatically. Human interpretations and agent proposals keep separate provenance.
Configure a reasoning step
Choose Choose next step from a card or its output. Pick Clarify, Explore, Challenge, Design a test, Compare or Synthesize. Disabled choices explain what the selection is missing. Challenge and Design a test require one exact claim or relationship target. A group frame organizes cards; select its actual material for the step. Compare needs at least two alternatives.
Write What should this step do?, fill its configuration and choose Make a step with these settings. This saves a draft without running a model or charging for a job. In the step inspector, review exact input revisions, supply missing typed inputs, approve human criteria for comparisons and keep personal context separate unless you explicitly include it. Save configuration records changes. A changed input makes an older result stale without erasing its receipt.
Review a path before it runs
Use Run step or select configured operations and choose Run selected path. The Private path preview lists the ordered steps, exact material, coverage, provider, executor, spending limits and any blockers. Resolve missing inputs, cycles and dependencies outside your selection. Then choose Review exact trace and Start reviewed path. Editing the path or its inputs requires a fresh preview. Creating cards or connecting sockets never starts this run.
Reuse the structure privately
Select a saved operation path and choose Save as private workflow, name it and press Save private template. A template keeps structure while leaving private text, evidence, results, decisions and permissions behind. When reusing it, choose Preview template and Add fresh workflow, then bind fresh inputs and configure its unresolved steps. Applying a template runs no tool and imports no prior approval.
Bring in what
you already know.
Files, links and live data connections all become nodes. The workspace keeps every original exactly as it arrived, extracts what it can read, records how much it covered, and lets you correct the result without losing any version.
Drop a file
Drag files anywhere over the canvas. The overlay reads Drop it into the investigation. with the limit UP TO 10 FILES · 25 MB EACH. The same limits apply to the file picker and to pasted files. Each file becomes a card whose body reads Original received. Preparing extraction…, then extracting…, then understanding… if a model is connected. If something fails the card says Needs attention · open to retry.
| Format | What is extracted | Limits |
|---|---|---|
| Text · Markdown | The text, split into passages. | 25 MB per file |
| CSV · JSON | Rows and columns. The upload also becomes an immutable dataset version that experiments can use. | 2 MB, 5,000 rows, 50 columns per dataset version |
| The text layer, page by page. Scanned pages fall back to English OCR. | 250 native text pages, 8 OCR pages, 45 seconds in total | |
| DOCX | Paragraph text. | Bounded document parsing |
| XLSX | Cell values as they were last calculated. Formulas are not executed. | 20 sheets, 5,000 rows per sheet |
| PNG · JPEG | Metadata only. Reading the picture needs a connected vision-capable model and consent. | 4 MB, 4,096 pixels per side, 8 megapixels, static images only |
Anything else is preserved but not read: This original is preserved, but its format requires a text or supported document export.
Read what was extracted
Open a source card. The disclosure Source text & reading coverage starts with a coverage line such as 12 of 30 pages extracted · 3 read with OCR · Review coverage, or 48,120 characters extracted · Complete extraction. Passages are labelled Page 4, Paragraph 12 or Passage 3 so a citation can point back to them.
OCR results come with honest warnings: reading order is approximate, numbers and tables need review before computation, and a confidence score is a recognition score rather than a probability of being right. Pages with low confidence are listed by number.
To fix an extraction, open Correct this text, edit Corrected source text and press Save a correction. The toast reads Correction saved as a new extraction revision. Nothing is overwritten. Extraction versions (N) lists every revision, and human corrections are marked as such. If the source has been re-extracted since you started editing, the panel asks you to compare before saving.
Two more buttons: Retry extraction runs the reader again, and Choose the original again replaces the file. Understand this source asks your connected model to summarise it into Extracted facts; the toast reads Source understanding queued.
Add a folder
Press + and choose Add a folder. The panel Add a folder. starts with Choose folder; your browser asks which folder to open. Nothing is uploaded until you confirm. The list shows every file with its size and preselects the eligible ones: up to 25 MiB per file and 250 MiB per import, with sensitive and generated files excluded. Use Select eligible files or Clear selection, then press Import selected files. Each file becomes a private source copy on My canvas, and the result line counts what was uploaded, already present or not uploaded. Extraction continues on each source card. Imports are copies, not a live connection; choose the folder again to import updates, and identical copies are reused. Imported copies may be sent to your chosen model provider for processing.
The same panel has Advanced: live folder access and local execution. It pairs a small companion program on your own computer (macOS, Node.js 24 and Python 3) with this investigation: Download folder companion, then Copy Connect command and run it in your terminal. You approve the computer in the browser and review the folder's included and excluded files on your computer. A folder card then shows its freshness and last snapshot; Inspect contents picks files to use, and answers cite the exact file version they read. Experiment in a working copy runs a reviewed command on captured files in an isolated local container with no network, after you approve that exact plan on your computer. Disconnect stops future access and keeps evidence you already imported.
Bring a link
Paste a web address on the canvas, or use Bring a link. The card is titled A source worth exploring and reads Reading this source… while the server fetches it. Public reads are bounded to 2 MB and 25 seconds. A server without web reading says so: Web reading is unavailable on this server. Add an excerpt or upload an export.
Connect an API
A data connection is a bounded GET request that uses the source's own credentials and turns each response into a dataset version. Its panel is called Connect it where you use it. and has four sections.
| Section | Fields | Notes |
|---|---|---|
| Request | Service name · Endpoint · Documentation · Authentication (No authentication, Bearer token, Named header key) · Header name · Data-source key · GET parameters | Parameters are typed string, number or boolean. The key is encrypted on the server and never shown again; the field later reads Replace source key (leave blank to keep). |
| Read policy | Refresh cadence (Manual reads, Every 15 minutes, Hourly, Every 6 hours, Daily, Weekly) · Reads per day (max) · Pagination (Single call, Page number, Row offset, Cursor from response) · Calls per read (max) · Rows per call · parameter names | Reads per day is 1 to 96, default 24. Calls per read is 1 to 25. Every read stops at 2 MB or 5,000 rows. The footer counts Reads in the last 24 hours. |
| Extraction | Row path (optional) · Columns to keep · Keep only rows matching a rule | Rule operators: equals, is not, contains, greater than, less than, is not empty. |
| Role on the canvas | What is this data for in the investigation? · Observe rows · Compute on it · Compare versions | The purpose is read by the agent as your intent, never as evidence. With no uses ticked, the agent only learns that the connection exists. |
Save connection stores the settings. Connect & test performs one real read and reports Read completed in 1 call(s). A versioned sample is ready., or explains why it stopped early: call budget, row limit or size limit. The Retrieved sample shows the first twelve rows. Disconnect removes the credentials; earlier samples remain.
Source keys live apart from model keys. Both are encrypted at rest. Publishing an investigation never includes original files, connection settings or credentials, and a model only sees a data connection's rows when you tick Observe rows or Compute on it.
Find your way
around.
The canvas has a tool rail, a zoom control, an undo control, a selection toolbar that appears when something is selected, and a status bar. All of them float over the canvas, and the help panel behind ? repeats what follows.
| Tool | Key | What it does |
|---|---|---|
| Select and move | V | The default. Click selects, drag moves. Shift + click adds to the selection; Shift + drag draws a selection box. |
| Pan canvas | H | Drag anywhere to move the view. Holding Space pans temporarily in any tool. |
| Connect nodes | C | Choose two square handles to connect them. The status bar reads Choose two square handles · Escape to cancel. |
| Group nodes | G | Drag a selection box around cards, or tap them, to make a research group. |
| Add source or node | + | Opens the add menu of chapter 03. |
| Select loose pieces | Selects ungrouped, unconnected My-canvas cards that are ready for review; pending source placeholders stay out. | |
| Ask about selection | Sends the current selection to the assistant. | |
| Agent | Opens the agent rail of chapter 09. | |
| Open notebook | Opens the notebook drawer of chapter 10. | |
| Canvas shortcuts | ? | Opens the help panel, Find your way around. |
The zoom control offers Zoom out, Fit, Zoom in and Toggle minimap. Zoom runs from 15 to 200 percent; below 40 percent, cards switch to a low-detail overview so a large canvas stays legible. F fits everything into view.
| Keys | Action |
|---|---|
| ⌘ / Ctrl + S | Save the project, even while typing in a field. |
| ⌘ / Ctrl + Z · + Shift | Undo · redo. |
| Esc | Close the open panel, cancel a connection, return to Select. |
| Enter | Inspect the focused card. In the Group tool, Space toggles its selection. |
| Arrow keys · Shift + arrows | Move the focused card 5 pixels · 20 pixels. |
| Backspace · Delete | Delete the selected cards and edges. Locked cards stay until you unlock them; Undo restores a completed deletion. |
| Double-click | On a card: inspect. On empty canvas: add. |
| Paste | A web address becomes a source, text becomes context, files upload. |

The selection toolbar
Select one or more cards and a toolbar named Actions for the selection appears. It counts the selection, then offers Next step, Find connections, Ask, more actions and Clear selection. Under more actions are Group, Trace, Compare alternatives, Connect and other actions that fit the selection. Drag its grip to move it; arrow keys nudge it and Home resets it.
The … button also leads to All actions for the selection:
- Next step: open the guided next action for the selection. Chapter 09 explains the agent rail.
- Branch and compare: keep this interpretation as a baseline. Chapter 11.
- Keep in notebook: a notebook entry pointing back to the selection.
- Lock / unlock evidence: pinned cards cannot be moved or removed by accident, and their status reads Locked evidence.
- Exclude from exploration: the agent will not use these cards. Their status reads Excluded. An optional Dead-end reason records why in the notebook. When the selection holds excluded cards, Include in exploration brings them back; the saved research direction follows the change, affected research pauses until you resume it, and stale evidence stays stale.
- Group selected evidence and Share selected evidence.
Read a busy canvas
Relationship labels stay quiet until you hover or select. Selecting a card fades unrelated cards and wires; click empty paper to clear the focus. Each card initially shows only its strongest few wires, with +N more for the rest. Open Layers to choose stages, focus a direction or Show only a group. Under Relationships & details, Show relationship labels and Show all wires reveal the full graph. These settings change the view, not the saved investigation.
Open a card and press Trace the reasoning to follow its numbered chain and the explanation for each step. Group connectors count bundled relationships; opening one shows the underlying links. A research direction places its hypothesis, test and result in readable rows, with supporting evidence and settings wired into the test.
Name a concept without deciding it
On an Idea card, choose Name the concept, or use Name a concept for selected evidence. Describe who it serves, the problem, mechanism and possible failure; keep observations, inferences, assumptions, alternatives and missing evidence distinct. Select the individual cards it rests on. A frame is not evidence. An assistant draft becomes a named concept only when you name it yourself. Naming never approves a risk or establishes the benefit.
Research groups
A group is a frame with a name and a purpose. The assistant receives the group's meaning together with the evidence inside it, so asking about a group is different from asking about its members one by one. Press G and a guide appears: Group connected evidence, Drag around cards, or tap cards to select them., and a live count such as 3 selected. Press Make a group.

The group's own panel, Give this group a purpose., holds Group name and What connects this group?, with the placeholder A shared question, hypothesis, theme, or reason these sources belong together…. Save group stores it. Ask this group sends the group, its nested groups and its sources to the assistant. The In this group list lets you remove members, Add selected (N) adds the current selection, and Select cards on the canvas returns to the canvas to pick more. Ungroup removes only the frame: Keeps every node and its connections. Undo restores the group.

Undo and redo
The floating undo control and ⌘ / Ctrl + Z step back through canvas edits: moves, additions, deletions, connections, group changes. Undo is aware of revisions, so it never discards an update that arrived later from a source reader or the computation worker. What it cannot undo, the help panel says plainly: Completed research actions and provider charges remain in history.
The status bar at the bottom reads 5 nodes · 4 connections, then a hint that follows the active tool, then Export canvas, then the visibility of this project: Private investigation, Private working copy or Public snapshot.
Choose your
model source.
Everything so far ran without a model. Source understanding, the assistant, research and vision need a ready source. Choose BYOM with your own provider account, or a qualified Synatrail-managed route where offered. You decide what it may process and accept a spending limit.
Model access
Press your avatar, Profile and settings, and open Model access. Choose an intelligence source and give processing consent. BYOM offers your personal provider connections; managed mode lists only qualified routes and asks for no personal API key or USD provider allowance. If managed routes are unavailable, the panel says so and does not send work through BYOM instead. BYOM offers these providers:
| Provider button | How you connect | Models by task · planning / building / bulk |
|---|---|---|
| OpenAI | API key | gpt-6-astra / gpt-6.1-sol / gpt-6-luna |
| Claude API | API key | claude-fable-5-1 / claude-sonnet-5-5 / claude-haiku-4-5 |
| Gemini API | API key | gemini-3.1-pro-preview / gemini-3.8-flash / gemini-3.5-flash-lite |
| Claude Code · Pro / Max | Browser sign-in with your Claude subscription | claude-fable-5-1 / claude-sonnet-5-5 / claude-haiku-4-5 |
| Codex · ChatGPT subscription | Sign-in with your ChatGPT subscription | gpt-6-astra / gpt-6.1-sol / gpt-6-luna |
Connected providers show Connected. The platform assigns each task to the model tier it needs: planning and evaluation to the top model, building and testing to the middle one, gathering context and bulk work to the light one.

With an API key
Under Model providers, choose a provider and paste your key into API key. Tick the consent line, which is required: I give this provider approval to process my investigation sources for approved work. Press Save connection. Saving does not start a paid model call, and the toast reads Provider connection saved. You set no model and no dollar allowance here: the agent selects a model for each task, and you approve the price when each job starts. Keys are encrypted on the server and never shown again. Remove key deletes one.
The capability check
Before automatic work, do a check of the provider's task access. The check is real and paid, and it starts only when you press the button. Set Maximum provider charge for this check (USD) and press Do a check of provider access. The toast reads Provider access check completed. Until you run it, the panel says Test the model connection before starting automatic research.
With a Claude subscription
Choose Claude Code · Pro / Max. The panel explains that your Claude plan will handle ideas, source understanding and research, then states the consent: By choosing Connect Claude, you select Claude for this workspace and allow it to process sources you add, generate ideas and research your approved public queries. Press Connect Claude, then Sign in on Claude ↗ and approve access with your own account. If Claude gives you a code instead, open Claude gave me a one-time code, paste it into Code from Claude and press Complete the connection. The status strip moves from Waiting for Claude sign-in to Claude connected. Your sign-in stays in a private hosted session for your next visit. Disconnect Claude ends it.
With a ChatGPT subscription
Choose Codex · ChatGPT subscription. The panel reads Use your ChatGPT subscription for Codex in Synatrail. Sign in to your account. Give approval to use your plan. Your plan limits and credit settings are applicable. Press Continue with ChatGPT; doing so allows Codex to process the investigation sources and work you approve. Sign in with your ChatGPT account. The status moves from Waiting for ChatGPT sign-in to Do a check of Codex task access to complete the connection. Press Do a check of Codex task access; the check uses a few requests from your ChatGPT plan, not an API key or platform credits. When the status reads Codex connected, choose Use Codex for this workspace. Disconnect Codex ends the connection. Cost lines show ChatGPT subscription · Cost unknown, because your plan, not Synatrail, bills the model work; plan usage limits still apply. On a server without this sign-in, the panel says ChatGPT subscription sign-in is not available on this server yet.
On a server where hosted sign-in is not enabled, the panel says Browser sign-in has not been enabled on this server yet.
Routes for answers and background work
For an explicit question, Provider chooses the connected intelligence source. The platform assigns models to their tasks; a provider choice does not establish capability or qualification. For the standing agent watch, a paired Claude Code connection can be verified yet offline. If it goes offline and you also have a verified API connection, background passes can use the lowest-cost verified API model until Claude Code returns; the rail names that fallback and charges it against the saved daily budget.
Tell the model about you
The Research context tab has two long fields, What are you trying to understand? and How should a model work with you?, up to 12,000 characters each. They accompany every question and mandate. Save research context stores them.
With BYOM, model usage remains on your provider account and autonomous work stays within its USD allowance or subscription limits. With Synatrail-managed intelligence, actual reported model tokens use the same platform credit balance as unchanged main-action fees, at the displayed model price based on actual tokens. Before work starts, a quote separates the action fee, first-call ceiling and one total run limit. Later calls reserve only what they need. Unknown usage remains a pending hold until reconciled; a quote is not a guaranteed final debit. Revoking consent pauses undispatched work, while already sent calls retain their original charge.
Talk to
your canvas.
The assistant dock at the bottom of the canvas answers questions about whatever you have selected: a card, a relationship, a group, or the whole canvas. Every answer cites the evidence it used, at the revision it saw.
Choose what to send
A scoped question uses the selected card, relationship or actual members of a group. Source capture fields What prompted this? and Why this caught my attention preserve personal context separately from extracted facts. In a configured operation, personal context enters the run only when you explicitly opt in. Source processing still requires consent for the chosen model source.
The dock
Its first line tells you the scope: Thinking with your canvas, Thinking with "A paper on urban heat", Thinking with 2 selected nodes or Thinking with the selected relationship. Next to it, Ask sends a scoped question, Steer tells the agent what to focus on within its saved limits, and Provider chooses the connected source for the request. The platform assigns the model route for its task.
The text field reads Talk to your canvas. Enter sends, Shift + Enter starts a new line. The book button on the first line opens the conversation history.
Without a model, sending shows To start research, connect a model and give source-processing permission in Model access. and opens the settings for you. With one, the toast reads Question saved. The assistant uses the selected material.


The conversation panel
Answers arrive in Think with the evidence. Messages are marked YOU and RESEARCH ASSISTANT. While the model works, a progress block titled Working with your evidence carries a LIVE badge, the latest status, and a disclosure Activity and model summaries that lists each tool call and summary. Text streams in under ANSWER IN PROGRESS.
A finished answer ends with three things. A disclosure counting its evidence references, each pointing to a card and revision. A disclosure called Limits and open questions. And a provenance line such as Selected model: OpenAI · gpt-6-astra. A Claude subscription run that cannot confirm which model actually replied is marked Actual model unverified.
Continue with Continue with this evidence and Ask a follow-up. If a run stops early the panel says so, The answer is paused. or This run finished without an answer., and offers Try question again, which retries with the original model and evidence, and View this run.
Keep an answer
An answer also proposes the evidence and sub-questions it calls for, as dashed cards on the canvas you asked from, marked Proposed · approve or reject. Their panel explains: The research assistant proposed this card with an answer. Give approval to keep it as your card. Reject it to remove it. Or reconnect it first. A proposed source, paper, study or dataset must carry a public address; the server checks the link before saving, and a card whose link could not be verified says so. For those cards the footer button on My canvas reads Give approval and read: approving also fetches the address as an ordinary source, with its extraction and coverage, so a proposal becomes evidence only once it has actually been read.
Every answer offers Add answer to canvas and Keep in notebook. Saving creates an Answer card connected to the evidence it cites. Open it to read Limitations and Evidence used, or choose Ask a follow-up. If the cited evidence changes, the answer warns you to review its original sources. The visible workspace is one living canvas; the agent's background map is not a second editing tab.
The edge panel's Interpret and Challenge buttons, the card footer's Explore, the group panel's Ask this group and the selection toolbar's Ask and Ideas all lead here. Make testable does not: it opens a test plan instead, in chapter 11.
Let curiosity
keep working.
The agent can follow your investigation in the background. It looks for useful connections, develops directions and tests them within the level and daily budget you save. Its cards appear on the working canvas, where you can inspect the evidence and make your own decisions.
Start the watch
Open Agent in the left command bar. The rail shows Discoveries, Activity, Decisions, Agent and Notebook. Under Agent, choose an automation level and daily budget, save them, then press Start the agent. The presence control reports Following this investigation and the analyses remaining today. Pause, Analyse now and Stop are available in the same rail; Stop asks you to confirm and leaves the map intact.
Manual acts when asked. Suggest keeps thinking and proposes next steps. Research can investigate and run private tests. Full auto can run supported, configured actions within the daily budget. The rail shows which abilities are available at the saved level and why an ability is unavailable.
Review a separate research branch
Where available, Investigate on a separate branch previews the saved goal, exact included and excluded evidence, approved public queries, allowed tools and allowance before Start this branch. The branch writes to the agent's private map. In branch review, select the additions or amendments you want, inspect their source revisions and conflicts, then apply the reviewed changes. Results remain proposals until that review; human decisions and commitments are not silently imported.
Approve public research
External research is off until you turn it on for this investigation. The rail shows the short, general terms that may be disclosed to public search. Review those terms and the search budget before saving the policy. Queries are composed from that approved policy; private notes and raw project text are not sent to public search. If a test needs web evidence and no policy is approved, its card says Needs you: Approve a search policy. You can approve one and let the test continue.
Read, steer and decide
Use Discoveries to read a candidate, then Explore, Why or Ignore. Why shows the source trail and any relevant notes the agent consulted. Activity shows recent passes and costs; Decisions records what you chose. In the assistant dock, choose Steer and write what the agent should focus on next. Steering uses the connected model and its saved limits.
The agent can remember hypotheses, dead ends, discoveries and retrieved sources across your own projects. Its notebook entries remain live references; they provide background for a new pass rather than silently deciding what is true on this canvas.
The Agent controls list the fees for jobs the agent starts by itself: Watch day 3 · Test plan 2 · Experiment run 3 credits. Charged only when the agent delivers a result to your canvas. A watch day is charged once per day, on a day the agent publishes a proposal or a notebook entry. If your balance or your daily credit limit cannot cover it, the watch keeps its free passes and says why in Activity.


Keep a thought.
Keep a finding.
The notebook is the investigation's memory: your notes, the agent's cited findings, saved answers, kept selections and experiment comparisons, each pointing back to the canvas it came from.
Open Notebook in the Agent rail, or choose it under More in the command bar. The drawer's overline reads COLLECT THE USEFUL BITS. Three filters: All notes, Agent findings and Yours.
Every entry is labelled Your note or Agent finding, with its kind, and carries a Return to canvas button for each card or edge it references. Agent findings offer Keep in my notes, which promotes them into your own notes so they survive a change of mandate.
Write your own with the field Something worth remembering… and Add a note. Other flows write here too: Keep in notebook from the selection toolbar and from answers, and Keep comparison in notebook from chapter 11.
Share notebook publishes the notebook alone. The canvas stays private. Chapter 13 explains the review step.
Make the connection
testable.
An experiment runs one of nine bounded numerical tools on your datasets and parameters. Inputs are pinned to exact versions when you run, results carry a receipt, and alternatives never overwrite the baseline they branched from.
Follow an agent test from proposal to result
On a research direction, the agent proposes a test and compiles a plan with a metric, threshold, disproof condition, sample rule, cost cap and method. The test card shows its state and Step 1 of 8 through Step 8 of 8: compiling, choosing data, reading sources, wiring, running, presenting, deriving and finished. A refusal or failed compile names the reason. At Research level a prepared test can run within the saved limits; at Suggest level you choose Run; at Manual level you can make it your own test.
The agent binds an existing dataset, extracts a cited public table, or asks for a value only you know. It makes separate evidence, context, setting and data cards and wires them into the test's four named bottom ports. Public evidence quotes must match a stored page word for word; the card labels the agent's reading as supporting or challenging, never as proof. The test panel has Plan, Data, Wired, Context, Receipts, Result, Agent reading, Derived and Your verdict sections.
A verified result appears beside the test with a measured number, unit, source and limits. It connects back to the hypothesis with an evaluates relationship. The hypothesis shows Verdict: supported by computation, Verdict: weakened by computation or Verdict: inconclusive. The agent then derives a candidate, alternative claim or sharper test from the result. You still record your own verdict separately.
Write your own test plan
Every Experiment card can hold a test plan: what the test asserts, what observation would show it is wrong, and how it is carried out. Make testable on a relationship opens one, already pointing at that relationship and the revision it has right now, with Open the relationship to go back to it. Choose A test you carry out yourself and the card is a plan and nothing else — nothing is queued and no calculation runs. Choose A calculation on your evidence and the calculation appears below; it still waits for Run experiment, and it will not run until the plan is saved, so the result names the exact plan revision it answers.
Changing the claim, the falsifying observation, the method or the target saves a new revision, and the heading counts it: TEST PLAN · REVISION 2. Earlier results keep the revision they answered. Writing down different conditions is not a new revision. If you later delete or rewrite the relationship the plan points at, the plan survives and says so.
Record what you observed
Once a manual plan is saved, Record what you observed appears under it. Write the date, what actually happened, the conditions as you carried them out, who did it and what the result cannot settle. Two questions are kept apart. Did the test run? is about the work: it ran and gave a clear observation, it ran but settled nothing, or it could not be carried out. Your reading of it is yours: not read yet, tentative, supports the claim, or refutes it — and a reading needs a reason.
The result is saved on its own Observation card, drawn to what it tested with an evaluates line, and written into your notebook, all in one save. Nothing is calculated and no receipt is invented. A result that refutes your claim still counts as a test you carried out, and it never changes the relationship's assessment on its own — you record that verdict on the relationship itself, with your reason. Saving twice never makes two cards. Correct this result saves the next revision of the same record and leaves the earlier entries in your notebook.
What a finished calculation tells you
When the worker finishes, the result appears as its own card in the investigation, labelled Computed result or Simulation result. That label comes from what actually ran, not from what anyone wrote, so a plan can never look like a finished output and a simulation always says it is conditional on its assumptions. Open the card to inspect its receipt; nothing you type can create or change one.
Open the card and you can read the exact versions it was computed from: the output version, the software and method versions, each frozen dataset with its hash and the moment it was saved, the bound parameters and any random seed. From there, Open the test plan takes you to the exact plan revision it answers, and the plan's Open the relationship takes you to the claim itself. The card says whether the run started from that plan or was linked to it after the run, so the order of events is never blurred.
Connect to what it tested draws an evaluates line from the result to the card it tested. It is drawn unassessed on purpose: accepting a result never makes a relationship supported. A run that failed or was cancelled publishes no card at all, and an unrelated result answers nothing — the plan tells you how many results answer its falsifying test as it states it now.
Open an experiment
Use Create an experiment in the add menu, Adjust on a parameter, Configure test on an experiment, Open result on a saved result, or Branch and compare from a selection. The panel is called Make the connection testable. Give it an Experiment title, choose a Calculation, and pick an Input dataset from the versioned datasets on this canvas: uploaded CSV and JSON files, and data connection samples. Without any, the panel says Drop a CSV on the canvas to make a versioned dataset available here.
| Calculation | Inputs | What you set |
|---|---|---|
| Profile a dataset | 1 dataset | Nothing. Returns row and column statistics. |
| Filter rows | 1 dataset | A column, an operator (equals, does not equal, is greater than, is at least, is less than, is at most, contains text, is one of, is missing) and a value. |
| Join datasets | 2 datasets | First join key, Second join key, Rows to retain (Matching rows or All rows from the first input), Trim key whitespace, Ignore key case. |
| Aggregate a measure | 1 dataset | Up to three Group by columns and up to eight measures, each a Measure (count, sum, mean, min, max, median, stddev) of a Measure column with an Output name. |
| Check or convert units | 1 dataset | A Column, a Target unit and whether to Convert compatible values. |
| Align time series | 2 datasets | Two timestamp columns, a Tolerance · milliseconds and a Direction (nearest, backward, forward). Use ISO dates or timestamps with a timezone. |
| Explore an association | 1 dataset | Two measures and a Method: Pearson · linear association or Spearman · rank association. Descriptive only, never causal. |
| Sweep a parameter | none | A Formula, an Output unit and variables, each with a unit and Values to compare · separated by commas. Up to 1,000 scenarios. |
| Simulate assumptions | none | A Formula, an Output unit, variables with Minimum, Most likely and Maximum, plus Draws, Random seed and an Outcome threshold. Independent triangular assumptions; percentiles describe scenarios, not forecasts. |
Units are checked first
Every unit field offers the supported symbols: 1, %, m, km, cm, mm, s, ms, min, h, day, kg, g, mg, K, °C, A, mol, Hz, N, Pa, J, W, kW, kWh, EUR, USD, GBP and count. Symbols combine with * and / and integer powers, so EUR/h and kWh/day work. Formula tools state it under the form: Formula units are checked before execution. Incompatible units stop a run before it starts, and a parameter with the wrong unit cannot be bound.
Bind a parameter
Every scalar field has a second dropdown, Bind … to a parameter, with Typed value and every parameter card on the canvas; incompatible ones are marked. Connecting a parameter’s round output to an experiment’s Settings diamond, or dropping the parameter onto the experiment, opens Put this input to work. Choose the Calculation input and press Bind input, or Bind in an alternative to leave a finished result intact and start a comparison branch instead. The Bound parameter values disclosure lists each binding with the note canvas value pinned at run.
Run, read, trace
Press Run experiment. The run passes through queued, running and completed, and a Computed result card appears in the investigation. In Saved experiments each result shows metric tiles, a chart, a table of up to twelve rows, and for simulations the disclosure Which assumptions move the result?
The Inputs & calculation receipt disclosure is the provenance: each input dataset with its version timestamp and content hash, each step with its method version and duration, the software version, the seed, and the Output version. Locate source jumps to the dataset card. Rerunning with unchanged inputs reuses the output and says so. Export result downloads everything.
Compare a change
Compare a change opens What changes? Choose a Saved baseline, name the alternative (the placeholder suggests What if the demand increases?) and press Branch & change assumptions. The note explains what happens: The alternative starts with the baseline's pinned datasets, parameters and calculation steps. Your canvases and baseline result stay intact. Change what you want in the branch and press Run alternative.
Then, under Compare saved outcomes, pick the Alternative result and press Compare changes. The verdict is one line, The output changed or The output is unchanged, followed by tiles for Inputs, Parameters, Recomputed steps and Unaffected steps, and a table of Measure, Baseline, Alternative, Change. Keep comparison in notebook records it. Preserved branches lists every branch with an Export branch button.

Show the actual test input inspector


A calculation may use 4 inputs, 16 steps, 5,000 rows, 100 columns, 24 parameters and 1,000 sweep scenarios, and must finish within 8 seconds and 96 MB. These nine tools do not execute generated code. Agent-run tests follow the saved automation level, budget and the compiled plan's own cost cap.
Since you
last looked.
Come back after a week and a banner above the canvas tells you what happened, why research stopped and what is still open. You do not have to read the whole log.
What it is made of
Only saved events: a source that arrived or was read, a correction, a finding or relationship the agent proposed, an answer, a calculation that finished and which plan revision it answers, your decisions and notes, and how a research run ended and why. Nothing is asked of a model to write it. Every line names the thing it came from and takes you there. If that card has since been deleted, the line still reads and takes you to Activity instead, which keeps everything.
Got it, and what it means
Got it marks what you just read as seen. That memory lives in this browser only — sign in somewhere else and the summary starts fresh. If more happened than fits on one page, the banner says so (Oldest 40 of 55) and Got it marks only the page you were shown; the rest waits for you next. Nothing marks itself seen on its own: leave the tab open in the background all week and everything that arrived is still there when you look.
Still open
Underneath, Still open lists what nobody has settled: what a stopped run said it could not answer, the evidence a finding says it is missing, and any test plan with no result yet. Answer drops the question into the ask box as a draft. Use as next goal puts it into the research mandate. Neither sends anything or spends anything — you still choose the model, the scope and the budget, and press Start investigation yourself.
Nothing is lost
by accident.
The canvas saves itself. When it cannot, it keeps your edits in the browser and tells you. When two versions disagree, it shows both and lets you decide field by field.
| State | Meaning | What to do |
|---|---|---|
| ✓ Saved | Every canvas change is on the server. | Nothing. |
| ○ Saving | A save is in flight. | Wait a moment. |
| Unsaved | Project name or question edited but not yet saved. | Save project or ⌘ / Ctrl + S. |
| Offline | The server is unreachable. Edits wait in this browser and retry every 8 seconds and when the connection returns. | Keep working, or Export pending edits. |
| Conflict | Another saved revision changed the same fields. | Examine changes. |
| Needs attention | The server rejected a save. | Open the conflict panel and Retry save or export. |
Closing the tab with unsaved work triggers the browser's own leave-page warning. If the browser refuses to keep an offline copy, a bar warns This browser could not keep an offline copy. Export your changes before closing. with an Export pending changes button.
When two versions disagree
A conflict shows a banner: Edits conflict with another saved revision. and 2 fields need a decision. Both versions remain available. Press Examine changes to open Keep your work. It lists each disputed field, marks entities that were removed in one version, and shows Your pending version and Current saved version side by side. Choose Keep my conflicting edits or Use the saved version. Everything that did not conflict is merged either way.
Undo works across all of this. It steps back through your own edits while preserving updates that arrived from a source reader, the computation worker or the agent in the meantime. The help panel is explicit about the boundary: Completed research actions and provider charges remain in history.
Share a reviewed copy.
Test one claim in public.
Nothing becomes public by itself. When you publish, you review the exact copy first. Other signed-in users can read it, vote, discuss, make their own working copy and send you changes. You accept or decline, and decide when the public copy updates.
Test one claim on the open web
On an eligible card, Start a public test for this claim prepares a landing page with one button or a waitlist asking for email. Fill the one claim, ask, audience, sender and a 1–90 day window. Preview shows exactly what leaves the investigation; fix any refusal before choosing Publish to the open web. The test page is reachable without signing in. Publishing a test and sharing an investigation are separate actions.
The public test shows responses and measured exposures where available; a missing denominator stays Not measured. Use Take it down and confirm to withdraw it early. After it ends or is withdrawn, Record the result keeps an Observation card and notebook trail. A response count is evidence of that test's outcome, not validation of the product's benefit.
Publish
Press Share in the header, or Share notebook inside the notebook. The panel is called Share your investigation. and starts with your status: Your investigation is private. or A public copy is available.
- What would you like to share?: Investigation · My canvas or Notebook only.
- A short introduction, required: What are you investigating, and why does it matter?
- What would help?: A source, a counterexample, a way to test an idea…
- Topics · separated by commas.
- Include my notebook entries.
Press Examine public copy. The review is the exact copy, marked REVIEW BEFORE PUBLISHING, with every card, relationship and included note, and one disclosure: the copy is visible to signed-in users, the source text shown is included, and original files, connection settings, credentials and your Agent canvas stay private. Press Publish reviewed copy. Later edits stay private until you return here and press Publish this update.
Make private asks for confirmation, Make this investigation private?, and warns that working copies others have already made remain theirs. Withdrawing blocks new reads, votes, comments and submissions.
The public board
Public investigations in the header opens the board, titled Public investigations. Filters: All investigations, Yours, Contributed to. A search field, Find an investigation…, and a sort, Most upvoted or Newest. Each row shows the vote control, the author, the title and introduction, topics, counts of comments, contributions, nodes and notes, the author's LOOKING FOR line, and Open investigation or Open notebook. Every signed-in user has one vote per publication, up or down, changeable any time.
Inside a publication
Four tabs: Canvas, a read-only copy you can pan and zoom, with every card and relationship listed below it; Notebook; Discussion, where Your comment and Add to the discussion start a thread; and Contributions. Copy link copies an address that returns to this publication after sign-in. On your own publication the buttons read Open your original and Manage publication.
Contribute
On someone else's publication, press Make a working copy. You get a private investigation of your own with every canvas tool, and a banner: Your working copy of Mara's investigation, with Discussion and Submit changes. The header's Share button now reads Contribute. Add a source, a relationship, a note or a counterexample, then press Submit changes.
The panel Start with a useful addition. shows only the actual differences, counted as new nodes, edited nodes, removed nodes, new relationships, edited relationships, removed relationships, new notes and changed question. Explain them in What does this addition help explain? and press Submit reviewed changes.
Review what others found
As the author, open your publication's Contributions tab. Each submission has Examine these changes, a field-by-field Before and Proposed view, and two buttons: Accept contribution and Decline. Acceptance merges into your private original and preserves your unrelated edits and private source bindings; conflicting edits are flagged for review. The toast reminds you of the last step: Contribution accepted into your original. Publish an update to share it.
Threads, people, connections
The panel Think together. has four tabs. Shared threads holds evidence snapshots published from a selection, with Copy to canvas, Report and, for your own, Withdraw. People lists users who opted into a profile, with Request a connection where they accept it. Connections lists requests to Accept, Decline or Close connection. Share selection, reached from the selection toolbar, publishes a reviewed snapshot of selected cards, optionally with source excerpts, after Review selected material. Administrators see a Reports tab to withdraw reported contributions.
Commission a handoff
Choose the exact evidence, review the planned outputs and price, then produce a versioned handoff. The package workbench keeps drafts, checks and released archives separate so you can see what is ready and what still needs a decision.
Open Produce package
Press Produce package in the header, or select material and choose Prepare a package draft. The panel opens Commission a handoff. This flow appears when comprehensive packages are enabled on your deployment. A deployment using the earlier composer labels the header Build package; those planning packages are described below.
Choose outcome, sources and delivery
Enter the Requested outcome. Select the Package type and Delivery stage. Read the explanation below each product feature. Select Yes, No or Not sure. A mixed product must include at least two product parts.
Select the material, notes, answers, concepts and decisions that the package can use. The package reads only the selected versions.
Select a Package level: Basic, Extended or Comprehensive. Read the list of files and the fixed price. Each level includes a specification, build plan, agent instructions, presentation and quality review. Higher levels add more design and operation files.

The package uses your saved AI settings. The platform selects and checks the models and file tools. You do not need to run a provider test or select individual tools. The review shows any missing product information or service problem before the package starts.
Review scope or buy a saved Pre-flight
Select Use only the selected material to make this package. Select Examine scope and quote to check the files, open questions and price. This review is free.
The optional Pre-flight — 5 credits saves a detailed plan. The plan explains the files, decisions, limits and checks. The full package starts after you accept its price.

Edit the inputs and use Examine the edited plan review. Revisions of that saved Pre-flight cost no additional Pre-flight credits. In the reviewed plan you can assign individual material cards to an artifact instead of using all selected material; selected notes and decisions remain part of its context. Changed inputs invalidate the previous start approval. The full run starts only after Accept … credits and produce package; managed model usage has its own reviewed credit limit.
Presentation delivery
Choose Include an editable PowerPoint (.pptx) with the presentation before reviewing the quote if you need editable slides. PPTX delivery requires qualified presentation rendering and checks. Read the presentation's captured views and output files in the artifact viewer; requesting a format does not bypass a missing renderer or failed validation.

Inspect the saved package
The Package workbench shows status, billing, artifacts grouped as Shared, Software, Service and Hardware, and a durable event timeline. Open an artifact to inspect its exact revision, authored source, rendered views, files and review evidence. Generated; review pending, Validated, Blocked and Stale are different states. Missing captured views and failed checks remain visible. Cancel and fence work stops further work; saved drafts and receipts remain.
Show the current package workbench

Edit and rerun
Where available, Edit canonical source as a manual revision saves an unvalidated source revision and invalidates dependent work. Edit derived exports through their authored source. Use Revise inputs and rerun for the full commission or Examine rerun of this artifact for a focused change. Review the new quote and reuse map. Valid unchanged artifacts keep their original bytes; affected outputs rebuild. Every rerun costs the full selected package price, and manual edits remain explicit source material for regenerated work.
Resume saved package continues eligible paused or blocked work. Resuming a cancelled or failed package requires acceptance of its saved spending envelope and a fresh build-release reservation. A manual draft uses Generate from fresh commission to review sources and capabilities before generation.
Freeze and hand over an exact release
Choose Freeze draft archive to preserve incomplete work. Freeze ready archive is available after completion and release checks. Release history shows the generation and manifest hash, with Download this exact archive and Copy agent handoff. The full archive must accompany the prompt. Give approval to use this manifest records your approval of that ready release; downloading does not approve it. A ready specification is not proof of physical safety, regulatory compliance or an implemented product.

Push a release to GitHub or Azure DevOps
Open Push this release to a repository on a frozen release. Review its draft or ready status, manifest, provider, repository, exact branch and unused package folder. Supply a one-time personal access token with repository write access, confirm I reviewed this release and authorize sharing it with this repository., then choose Push release. This sends the complete immutable package, including binary files, to people with access to that repository.

Show the Azure DevOps form

The token is cleared on submission and is used only for that request. GitHub needs Contents write permission and an initialized repository; Azure DevOps needs Code read and write. A new branch starts from the default branch; an existing branch can advance only from the exact commit reviewed for the export. Existing folders are rejected. If export history says pending or uncertain, enter a token and use Reconcile export to verify the remote files before starting another push.
Earlier planning packages and pitches
The walkthrough below describes the earlier bounded composer and the saved light packages used in the cookbook. Its single-turn planning archive and controls differ from Produce package. Historical records keep their original contents and provenance; importing one into the new flow creates an unapproved manual source rather than granting it validation or credit status.
Open the composer
Press Build package in the header, open Current project and press Make a build package under FROM INVESTIGATION TO BUILD, or select the exact cards you want and choose Prepare a package draft from the selection menu. The selection route keeps those cards pre-marked. The panel is titled From investigation to build. and explains the sequence: mark, review, research the pitch, then build.
Mark what belongs
Three lists with All and None buttons: CARDS · 8/8, NOTEBOOK · 3/3 and EARLIER ANSWERS · 2/2. Cards you had selected on My canvas are pre-marked; excluded cards, pending proposals and group frames are not offered. Earlier answers show the question they answered. At most 30 cards may be marked; the form says Mark at most 30 cards. Group or unmark some before composing. Then choose What kind of product: Let the evidence decide, Software, Hardware or Hardware and software; write Steering (optional), whose placeholder suggests Who it is for, constraints, what must be left out, preferred stack or parts…; and pick the Model provider. Three buttons close the form: Plan the pitch research, Do a scope check with the agent and Skip the pitch, build the package. Without a connected model, any of them takes you to Model access.
Check the scope with the agent
One bounded turn over the whole investigation. The review names the unmarked cards, notebook entries and answers the package would need, as RECOMMENDED · 2 UNMARKED ITEMS · 1 ESSENTIAL, each with a reason and a checkbox, and lists GAPS THE INVESTIGATION DOES NOT COVER with the warning Gaps become open questions in the package unless you add evidence first. Then: Add 2 and plan the research, Plan with my marks only, Skip the pitch, build the package or Back to marking.
Research the pitch
Planning is one more bounded turn that proposes up to eight generalized public search queries, each with what it verifies. The review says The agent will send general queries to public search only after your approval. Before you give approval, examine the queries. Change or remove a query that can show private information. Then give approval for the remaining queries. Edit any line, untick any, or Add a query. Set the limits: Searches (1 to 8), Sources read (1 to 12), Minutes (3 to 30) and, for API providers, Max USD. Press Start the research · 5 queries. The panel then reads Researching the pitch with public sources… with live counts of searches, sources read, turns and USD, and a log of every SEARCH and READ. You can close it; the pitch appears in the idea library when it is ready.
| Block | What it shows |
|---|---|
| Verdict | Pursue, Reshape first or Reject, with CONFIDENCE as a percentage, and an EVIDENCE line such as 5 of 7 claims cite a read source · 4 sources read · 5 searches. |
| Four meters | Desirability, Feasibility, Viability, Evidence, each 0 to 100. |
| Value proposition | FOR, PROBLEM, PROMISE, EDGE. |
| Market fit · Market size | Segments with a five-dot fit, WHY NOW, and TAM, SAM, SOM bars with a low, medium or high confidence chip and their basis. |
| Positioning · Competition · Whitespace | A two-axis map with This idea among named competitors, each competitor's approach with a + strength and a − weakness, and the open opportunities. |
| Reasons to reject · Concerns | Each reason with a severity chip (fatal, serious, minor) and a mitigation; concerns by area. |
| Assumptions · Unknowns · Could not verify | What the pitch rests on, what it could not settle, and claims for which no read source was found. |
| Sources read | Every public source the agent actually read, numbered; citations in the pitch link to them. A claim without one is marked unverified. |
Under the pitch: Build the full package, Revise the pitch, which returns to marking, Download, which saves the pitch as Markdown, and Remove. A pitch cannot be sent to a build tool; only a package can.
The package
Composing the package is one large bounded turn. The panel reads Composing the build package from your marked evidence… and warns that it can take a few minutes. When it is ready the panel is titled A product, defined. Its head states HYBRID · 3 modules · 7 stories · 2 parts · 1 drawings, the product name and one line. Four tabs follow.
The earlier package viewer also shows frozen specification status and a Requirements tab. Requirements keep their investigation basis and human review separately from generated prose. Review drafts, request review and approve exact requirements when available. A final Give approval to use this revision for the handoff records approval of that revision; exports never imply approval. Historical records without complete lineage remain labeled as historical.
- Overview: Problem space (who, pain, context), Solution idea with its differentiators and NOT IN SCOPE, Architecture, Open questions, the separately frozen Recorded investigation basis with exact selected material, named concepts, next test, assumptions, unresolved risks and selected human decision notes, and Evidence, where each cited card is a button that jumps to it on the canvas. Provider prose cannot replace the recorded basis, and the viewer does not infer that a nearby decision chose a concept.
- Build plan: Modules with dependencies, Waves marked parallel or sequential, PRDs as one disclosure per module with numbered stories and their acceptance criteria, Bill of materials for hardware, Agent concurrency, Controls and tests and Dependencies.
- Drawings: the physical parts of a hardware or hybrid package. Chapter 16.
- Full spec: the whole definition as Markdown, the same text that ships as PACKAGE.md.
Every list in a package has a hard bound, so it is a definition, never an open-ended document: at most twelve modules, six waves, eight stories per module, thirty parts, twelve physical parts, eight risks, eight open questions.
Send it to a build tool
Under SEND TO A BUILD TOOL: VS Code downloads the archive; unzip it, open the folder and start your coding agent, and AGENTS.md tells it where to begin. Lovable, Base44 and Replit Agent copy the build prompt to your clipboard and open their new-project page; a disclosure Build prompt for Lovable · copied keeps the prompt in view. The prompt embeds the spec up to 24,000 characters and points to the archive for the rest. Download archive and Remove sit at the bottom.
| File | Contents |
|---|---|
| PACKAGE.md | The complete definition plus the frozen recorded basis: problem space, solution idea, concept, data, architecture, modules, build waves, PRDs, bill of materials, geometry, design system, agent concurrency, controls and tests, dependencies, risks, open questions, evidence, selected concepts and materials, next test and retained human decision notes. |
| AGENTS.md · CLAUDE.md | Instructions for the coding agent: order of work by wave, concurrency rules, definition of done, and where the CAD source lives. |
| prd/<module>.md | One PRD per module with its stories and acceptance criteria. |
| BOM.csv | Parts, quantities, specifications, estimated cost and supplier hints, for hardware and hybrid products. |
| package.json.txt | The same definition and separately frozen recorded basis as structured data, with the run and receipt they came from. |
| cad/ | A README of checks and one folder per physical part with source and exports, once drawings exist. Chapter 16. |
The idea library
Press Projects and choose Idea library in the sidebar. Every pitch and package you composed is a card: PITCH with its verdict and confidence, or the package kind with 3 modules · 7 stories · 2 parts. Each offers Open, which switches to the package's investigation, Download, a Send to… selector for packages, and Remove. Before anything is composed the page reads No pitches or packages yet.
Source review, research and generation use the source and quote you select. BYOM model calls bill your provider; qualified managed model usage bills separately in platform credits. The earlier scope check and query plan use one bounded turn each, pitch research stays within approved search and spending limits, and its planning package uses one large turn. Produce package has its own reviewed fixed package price and optional 5-credit Pre-flight.
Read native CAD and drawings
Hardware outputs depend on package level, declared parts and qualified native tools. In Produce package, open the relevant hardware artifact and inspect its source revision, rendered sheets, measurements, checks and downloadable files. Comprehensive packages can include per-part CAD, drawings, PCB and fabrication outputs where applicable. Missing native or reviewer capability blocks generation; a rendered view alone does not establish dimensional or electrical correctness.
Open an interactive 3D preview
Where an artifact supplies GLB or STL, its CAD preview offers Model file and Open 3D preview. Drag to orbit, right-drag to pan and scroll to zoom, or use Fit model, Isometric, Front, Top, Right, rotation and zoom buttons. With the viewer focused, arrow keys rotate, Shift + arrows pan, + or − zoom and Home fits. Close 3D preview releases the viewer; downloads remain available if WebGL is unavailable. Use source CAD and validated measurements for manufacturing dimensions.

The earlier Drawings tab
The instructions below apply to physical parts in earlier planning packages. Each part gets a bounded source-writing turn and an isolated runner job. Mechanical exports can include STEP, STL, GLB, DXF and dimensioned SVG views. The connected runner's tool list determines availability, including whether electronic parts can use KiCad.
The parts in the package
Section 10 of the spec, Geometry, lists up to twelve parts, each with an id, a kind (mechanical or electronic), an envelope in millimetres, a material, a mass, the module and bill-of-materials line it belongs to, and a preferred tool: build123d for machined, moulded and printed parts, openscad for simple primitives, kicad for a schematic. Software packages have no parts and no Drawings tab.
The Drawings tab
The tab opens with a line naming what the runner can do, RUNNER · BUILD123D · OPENSCAD, and one entry per part: its name, MECHANICAL · BUILD123D, its stated 100 × 60 × 25 mm · PLA · 40 g · module firmware and description, and a status chip. Press Generate drawings. The chip moves through Writing the source…, Queued for the runner and Building on the runner to Drawings ready when every check passes, or Checks failed or Failed with the reason. Generation needs a connected model and the package's own investigation open; otherwise the tab says Open this package's investigation to generate geometry. A deployment without a runner says No geometry runner is connected to this deployment. and keeps the button disabled.
What comes back
Four views, FRONT, TOP, SIDE and ISO, each with a title block reading 1:1 · mm · third angle and overall dimension lines; hidden edges are dashed. A table of checks with PASS or FAIL in the last column. A line such as build123d · attempt 1 · 0.11.1 · 5.1 s · about 74.4 g. The files: source.py, source.scad or source.kicad_sch, part.step, part.stl, part.glb, top.dxf and the four SVGs, each a download with its size. Two disclosures, Engineer’s notes and Assumptions, hold what the model said about its own work.
| Check | Passes when |
|---|---|
| Solid geometry | The volume is above zero and there is at least one solid. |
| Watertight mesh | The exported mesh has no open edges. Shown only when it fails. |
| Envelope matches stated dimensions | The measured bounding box matches the stated envelope within 5 percent and at least 0.5 mm per axis. The axes are sorted first, so orientation does not matter. |
| Mass from volume and material | Volume times the material's density lands within 20 percent of the stated mass. The density comes from the model's statement or from a table of common materials. |
| ERC clean · Components placed | A schematic has zero electrical rule errors and at least one placed component. Electronic parts need KiCad on the runner. |
Revisions
When BYOM is selected, a failed build or check can trigger one included revision turn with the previous source, checks and runner log. Managed revisions require a separate spending review. Later versions keep their earlier history; Revise with the failures can request a further attempt within the part limit, while Regenerate starts a new build for a passing part.
In the archive
Once a part has completed geometry, the package archive gains cad/README.md, with each part's tool, verdict, measured envelope and mass and every check, and cad/<part>/ with the source and exports of the latest completed version, preferring one that passed. AGENTS.md gains a CAD section telling the build agent to work from that source instead of redrawing, and the build prompt for hosted tools mentions the folder.
Twelve parts per earlier package. Source of at most 60 kB; generated CAD runs only in the isolated runner with its allowed tools and limits. A job has up to 170 seconds and 25 MB of output, two runner attempts and three attempts per part. Available tools are reported by your deployment; electronic outputs require qualified KiCad capability. Passing geometry checks does not certify a physical product.
Each drawing run is one model turn on your provider and one job on the geometry runner. The automatic revision after a failed check belongs to the same job. A drawing job has a credit price shown before you start, and an automatic revision never costs a second job.
Your profile
and your account.
Your avatar opens the panel Your workspace. Besides Model access and Research context, it holds your profile, your account and, for administrators, the member list.
Profile
Display name, Bio, Skills, Interests and Your current question describe you to other users. Two checkboxes decide visibility: Show my profile in People and Accept connection requests. Both are off until you turn them on. Save profile stores them. Your avatar shows the initials of your display name.
Privacy choices and cookie consent
In Account, open Privacy choices to save or withdraw optional product news and pseudonymised platform research separately. Declining either keeps your account and investigations available. The page shows the recorded choices. These permissions are distinct from model-processing consent in Model access.
Use Privacy settings on the website or app to decline, allow or customize optional analytics, or download this browser's consent receipt. Analytics remains blocked until your current choice is saved and verified; withdrawal stops further collection. Optional analytics uses generic screens and allowlisted actions rather than project titles, source text, prompts or private workspace addresses. Cookie choices do not authorize model calls or sharing an investigation.
Account
Export workspace downloads everything you own as JSON. Below it, Current password, New password and Change password & sign out. Sign out is refused while edits are pending: Resolve or export your pending edits before signing out.
Credits · available now
The Credits tab shows your available balance, a This week summary of what agent jobs cost since Monday, the packs you can buy and a history of every reservation, charge and refund, each named by its job. Commissioned jobs show their fee before you start: a research run, a researched pitch, a build release and a drawing job. Small agent jobs cost one to three credits and are charged only when their result reaches your canvas: a test plan 2, a watch day 3, an experiment run 3, a critique 2, a card operation 1 (Synthesize, Compare or Challenge) and a step re-run 1. Buttons that start them show the fee, for example Run step · 1 credit. Your own work on the canvas is never charged. With BYOM, your model provider bills through your own key or plan; Synatrail-managed model usage is charged separately in platform credits. The pricing page explains what is never metered and what credits buy.
Payment method
Paid work needs a saved payment method. In Profile, the Payment method section shows No payment method is saved. or A payment method is saved. Add payment method opens a secure hosted payment page; adding a method makes no charge, and the status confirms Payment method saved. No charge was made. Manage payment method changes it later, and Refresh payment status reads the latest state. Without one, a paid job opens a short notice that links to this section; your credits stay separate.

Appearance
The header's Change color theme button switches between the sand-paper light theme and the deep blue dark theme. The choice is remembered in this browser, and the sign-in page and the website follow it.
Administration
Administrators can open Consent audit to inspect consent records and policy versions. Account choices and browser cookie receipts remain separate.
Administrators see one more tab. Invite an email address (optional) and Create invitation produce an invitation code, and the member list shows each email, name, organisation, role and status with Suspend and Activate. For an active member, Create password reset link shows a link that works once, for 30 minutes; send it to that person yourself. Registration mode is set by the operator. On synatrail.com it is open: the sign-up form admits and signs in anyone who completes it, and invitation codes are only needed on closed deployments. The details you gave at sign-up, your consent and its version are kept with your account, as the Terms of Use and Privacy Notice describe.
On a company installation that runs on your own servers, Administration also shows Company licence: the licence term, the allowance, the packs, the credits used this term, and the remaining credits or any overage. All users draw from that company pool; work does not stop when the allowance runs out. Below it, Usage reports lists the one credit-usage report the installation sends Synatrail each day, exactly as it was sent. A report carries usage figures only, never your content.
Addresses you can bookmark
| Address | Opens |
|---|---|
| …/app/#canvas | The current project's canvas. |
| …/app/#community | The public board. Add / and a publication's identifier to open one publication. |
| …/app/#setup | Settings, on the Profile tab. #setup/models opens Model access; #setup/account opens Account. |
| …/app/#people | The People tab of the Think together. panel. |
| …/app/#commons | The shared threads panel. |
| …/app/terms | The Terms of Use and Privacy Notice, also linked from the sign-in page. |
| …/app/privacy/preferences | Optional account choices for product news and platform research. |
Words, numbers,
and what is not here yet.
- Investigation · Project
- One question with a working canvas, sources, notebook, experiments and research history. Private unless published.
- Working canvas
- The shared view of your cards and the agent's developing cards, with provenance and tentative states visible.
- Agent map
- The agent's background record of hypotheses, directions, tests and sources. Its useful work is woven onto the working canvas for inspection.
- Node · Card
- One item on a canvas, of one of the kinds in chapter 03.
- Relationship · Edge
- A typed, directed connection with a rationale and an assessment. A hypothesis, not proof.
- Source
- An uploaded original or a public web page, with its extraction, coverage and revisions.
- Dataset version
- An immutable table made from a CSV, JSON or data connection read, identified by a hash.
- Data connection
- A bounded GET request with its own credentials, read policy and role on the canvas.
- Parameter
- A card holding a value, a unit and a range. Bound to experiments; pinned at run.
- Publication
- A frozen, reviewed copy of My canvas and optionally the notebook, visible to signed-in users.
- Research group
- A frame with a name and purpose around related cards. The assistant reads the purpose and the members.
- Mandate
- Goal, scope, constraints, approved public queries and limits given to the agent.
- Agent idea
- A tentative card or relationship developed by the agent. Its provenance and evidence remain inspectable; your verdict is separate.
- Receipt
- The record of one model or computation action: content, usage, cost and evidence used.
- Allowance
- The ceiling in dollars you set for a run on your own API key. Your provider’s money, reserved before a run starts and settled from receipts.
- Platform credits
- Prepaid units in one balance. They pay the displayed main-action fee for a research run, pitch, build release or drawing. When qualified Synatrail-managed intelligence is available and chosen, separately priced model usage based on actual tokens uses these same credits. BYOM model charges remain with your provider. Buy credits in the Credits tab.
- Harness · Articulation
- The software around your chosen model source: context, tools, bounds, memory, evidence checks, experiments and packaging. It turns model answers into work you can check and build from. Your model source and its charges are shown before work starts.
- Baseline · Alternative
- A completed result, and a branch from it with changed assumptions. The baseline is never overwritten.
- Working copy
- A member's private investigation made from a publication, used to prepare a contribution.
- Contribution
- The actual differences between a working copy and the publication, submitted for the author's review.
- Pitch
- A researched verdict on an idea: value proposition, market fit, competition, whitespace, reasons to reject, every claim citing a source the agent actually read.
- Build package
- One bounded specification handoff from selected evidence, with versioned artifacts, checks and frozen releases. Earlier packages contain a light product definition and planning archive.
- Configured step
- A saved reasoning operation with exact input references, human intent and required capabilities. Configuring it does not execute it.
- Path preview
- The exact ordered operations, material versions, authority and limits you review before starting a path.
- Pre-flight
- An optional saved 5-credit explanation of a planned package. Its edited inputs can be reviewed again without another Pre-flight charge; full generation has a separate price.
- Release manifest
- The frozen inventory and hashes of an exact draft or ready archive. Approval and repository export refer to that release.
- Idea library
- The section of the investigation library that keeps every pitch and package across investigations.
- Geometry · Drawings
- Parametric CAD source for one physical part, built by the runner into STEP, STL, GLB, DXF and dimensioned SVG views and checked against the package.
- Runner
- The isolated container that executes geometry source. No internet, no data, only the source in and files out.
| Area | Limit |
|---|---|
| Uploads | 10 files at a time, 25 MB each. Originals are always kept. |
| PDF reading | 250 native text pages, 8 OCR pages in English, 45 seconds. |
| Images | Static PNG or JPEG, 4 MB, 4,096 pixels per side, 8 megapixels. Reading needs a vision-capable model. |
| Public reads | 2 MB and 25 seconds per page. |
| Datasets | 2 MB, 5,000 rows, 50 columns per version. Spreadsheets: 20 sheets, 5,000 rows each. |
| Data connections | 1 to 96 reads per day, 1 to 25 calls per read, 2 MB and 5,000 rows per read. |
| Computation | 4 inputs, 16 steps, 5,000 rows, 100 columns, 24 parameters, 1,000 scenarios, 8 seconds, 96 MB. |
| Research allowance | 0.10 to 100 dollars per connection, 10 by default. A mandate: up to 100 dollars, 60 minutes, 30 turns, 12 searches, 20 sources. |
| Earlier package limits | 30 marked cards, 80 notebook entries, 20 answers. The earlier composer uses one turn each for scope review, query planning and a 16,000-token package within 6 minutes. Pitch research: 8 queries, 8 searches, 12 sources, 30 minutes. Produce package uses its reviewed artifact plan and spending policy instead. |
| Geometry | 12 parts per package, 60 kB of source, 170 seconds and 25 MB of files per build, 2 runner attempts per job, 3 attempts per part. |
| Text fields | Titles 200 characters. Card bodies and notes 18,000. Mandate goal 6,000. Rationale 2,000. Package steering 2,000. |
Not in this release
- Simultaneous editing of one canvas by several people. Contributions go through working copies and review instead.
- Multi-factor or federated sign-in. Accounts are email and password, managed by the administrator.
- Arbitrary executable connectors. Data connections are bounded GET reads and numerical experiments use the nine supported tools. Native generation and rendering require qualified isolated workers.
- Photoreal renders of hardware parts. Drawings, not renders, for now.
- Native tools without qualification. CAD, electrical schematics, PCB outputs and editable PowerPoint depend on installed, qualified tools and the selected package scope. The reviewed plan names unavailable capabilities.
- Importing the JSON exports back into a project.
- Codex as a personal research companion route. Codex works through the ChatGPT subscription connection instead (chapter 07).
One last thing the workspace says about itself, and this documentation repeats: implemented features are not claims of verified scientific findings. The example in these pages is a way to learn the tools, not a result.