Realtime multiplayer collaborative canvases by Incrementic (lean.incrementic.com), powered by Next.js 15, Convex, and AI stress-testing. Choose a Lean Canvas or a GTM Canvas for each new canvas.
• Canonical Domain: https://lean.incrementic.com
• Brand System: Incrementic Design System (Sora, Inter, IBM Plex Mono, #EA5148)
Each canvas has one template, chosen when it is created and fixed after that. Existing canvases with no template are Lean Canvases.
Realtime multiplayer collaborative canvases by Incrementic (lean.incrementic.com), powered by Next.js 15, Convex, and AI stress-testing. Choose a Lean Canvas or a GTM Canvas for each new canvas.
Both templates use the same 7 stress-test scores, so the stored results and the riskiest-assumptions view work the same way. Only the labels and the prompt change. The GTM blocks are based on common go-to-market frameworks (ideal customer profile, positioning, channels, sales motion, pricing, and launch plan), since there is no single standard GTM canvas.
Deployment: Vercel (Git-connected to incrhst/leancanvas-live)
Connecting to Claude (No-Code Connector URL)
For non-technical users, connecting LeanOn to Claude takes just one click:
In Claude, go to Settings › Connectors (or Add MCP Server).
Click Add Connector.
Enter the hosted MCP URL:
https://lean.incrementic.com/api/mcp
Click Save / Connect. That’s it! Claude will automatically discover the tools.
Available MCP Tools for Claude
When configured, Claude has access to:
list_canvases(): Lists your canvases with their template, role, and URL.
create_canvas({ title, description, template, seedNotes, launchDate }): Initializes a new canvas (template: lean by default, or gtm) and returns the direct collaboration URL.
update_canvas({ canvasId, title, description, launchDate }): Changes a canvas's title, description or launch date. The launch date (YYYY-MM-DD) is day 0 of the plan; null clears it.
get_canvas({ canvasId, driverUserId, market }): Fetches full structured canvas, blocks, and current notes with evidence states, drivers and tests (measure, passMark, reviewDate, latestResult). With a launch date it also returns currentDay and each note's reviewDay. driverUserId (or "unassigned") narrows it to the notes one person is driving; market to the notes that hold in one market (tagged with it, or untagged, which means every market).
list_canvas_members({ canvasId }): Lists everyone with access (userId, name, email, role). Use a userId as a note's driverUserId (the driver is the person doing the work on the note; viewers can be drivers) or a decision's deciderUserId.
add_note({ canvasId, block, content, evidenceState, driverUserId, markets, measure, passMark, reviewDate, reviewDay, latestResult, reason }): Inserts a note into a block of the canvas's template. A block from the other template is rejected.
update_note({ noteId, content, block, evidenceState, driverUserId, markets, measure, passMark, reviewDate, reviewDay, latestResult, reason, link }): Rewords a note, moves it to another block of the same template, changes its state, or sets its test. Pass only what changes; null clears a test field. Dates are YYYY-MM-DD, or give reviewDay (e.g. 30) on a canvas with a launch date; latestResult is { text, date?, verdict? } with verdict pass, fail or inconclusive.
delete_note({ noteId, reason }): Removes a note. Its history stays readable.
get_note_history({ noteId }): Lists every change to a note, newest first: who, when, app or agent, from and to, and the reason. Works for deleted notes.
request_decision({ noteId, question, deciderUserId, dueDate | dueDay, defaultOutcome, options }): Asks one canvas member to decide on a note. They're emailed a link () that answers it from a phone with no account or sign-in, even with view-only access. is what happens if nobody decides by the due date ("If nothing is decided by 1 November, we sign the agency."); it is shown everywhere the decision is, and never applied for them. gives 2 to 4 named choices instead of approve / reject.
Access Control Matrix
Actor
View Canvas
Edit Canvas
Comment
Invite Others
Run Stress Test
Answer Decisions / Check-ins
Anonymous
Yes (via /share/[token])
No
No
No
No
Only with the decider's emailed link (/d/[token]), for that one decision. Can send a test result with a note's result link or a driver's check-in link (/r/[token]); it waits for an editor to accept it
Authenticated Viewer
Yes
No
Yes
No
No
Yes, when they're the named decider (decisions) or the note's driver (check-ins)
Authenticated Editor
Yes
Yes
Yes
Yes (Editor / Viewer)
Yes
Yes, when they're the named decider (decisions) or the note's driver (check-ins)
Owner
Yes
Yes
Yes
Yes
Yes
Yes, when they're the named decider (decisions) or the note's driver (check-ins)
Routes
/: Landing page with features & call to action
/login: Magic link authentication with automatic return URL redirect
/dashboard: Workspace canvases list & creation modal
/canvas/[id]: Full multiplayer collaborative editor with sticky note evidence cycling
/share/[token]: Public read-only view for anonymous users (editing and stress-test triggers strictly disabled)
/decisions: The signed-in user's open decisions, across canvases
/d/[token]: Public answer page for one decision, from the request email. No sign-in. The token is stored hashed, works for that one request only, and stops working once the request is answered or withdrawn, when an editor sends a new link or turns it off, or when the decider leaves the canvas
/r/[token]: Public result page for one note's test: what we watch, the pass mark, a box for the result (a number or a line), an optional "did it pass?" and name. No account or sign-in. Two kinds of link open it: a shareable link an editor gets from the note (copy, or share to WhatsApp or email), which takes results as often as the test needs until an editor makes a new one or turns it off; and one link per question in the weekly check-in email, which answers that question once (a result, or "nothing new") and stops. A result sent this way never changes the note by itself: it waits on the note, marked "via link, no account", and the operator is emailed. An editor accepts it (it becomes the latest result; the evidence state is still theirs to change) or rejects it. Each link takes at most 10 results an hour, and a note stops taking more once 20 are waiting
/invite/[token]: Magic invite acceptance flow (requires authentication before joining)
Convex Functions
canvases:createCanvas: Create canvas (optional template, default lean) and default workspace with owner membership
canvases:getCanvas: Secure query for authenticated members
canvases:getCanvasByPublicToken: Sanitized read-only query for public sharing
canvases:setPublicView: Enable/disable public read-only link and regenerate token (Owner only)
notes:addNote, updateNote, deleteNote, reorderNotes: Editor+ only mutations
answer_decision({ noteId, answer, choice, comment }): The named decider answers approve, reject, choose with one of the request's options as choice (all three make the note a decision), or change (needs a comment; sends it back with the note unchanged).
withdraw_decision({ noteId, reason }): Cancels an open request.
list_decisions({ canvasId }): Decision requests on a canvas, or, without canvasId, the open ones waiting on you.
create_snapshot({ canvasId, label }): Freezes the canvas under a label, e.g. Day 30.
list_snapshots({ canvasId }): The canvas's snapshots, newest first.
compare_snapshots({ canvasId, from, to }): Notes added, removed or changed between two snapshots; to defaults to current (the canvas now).
get_review({ canvasId, sinceSnapshotId }): The review-meeting view, also at /canvas/[id]/review: tests grouped by how the latest result compares with the pass mark (missed, no result yet, too early, met), overdue reviews, open decisions, and what changed since the latest (or given) snapshot.
list_check_ins(): The weekly check-in questions waiting on you.
answer_check_in({ noteId, hasEvidence, text, verdict, date }): No records "no new evidence"; yes needs a line, which becomes the note's latest result. Works for view-only drivers.
send_check_in({ canvasId }): Sends this week's check-in now. Otherwise a cron sends it every Monday at 13:00 UTC (8am Jamaica) to the driver of every note that is unproven (unknown, assumption, observed) or has a test, and isn't a decision.
run_stress_test({ canvasId }): Triggers the 7-dimension diagnostic (Ash Maurya for Lean, go-to-market criteria for GTM) and extracts top riskiest assumptions.
export_canvas({ canvasId, view, format }): Returns a link that downloads the canvas (view: "canvas") or its riskiest assumptions (view: "riskiest_assumptions", needs a stress test first) as a PDF or PNG (format). The file is rendered in the browser, so open the link while signed in. format: "summary" instead returns plain-language text right away: one short paragraph per block for a non-technical reader.
