What is canvai CLAUDE.md?

Repository instructions for Canvai, an open-source visual discussion board that uses JSON Canvas files to store diagrams and discussions. They define project rules, design decisions, documentation language, and data-preservation requirements.

Guiding changes to the Canvai server, web client, canvas files, documentation, and tests while preserving unknown data and keeping design decisions consistent.

They give coding agents the context needed to avoid putting private notes in the public repository or breaking the file format and layout rules.

canvai CLAUDE.md — Instructions file for Claude Code by chuck00lin: Repository instructions for Canvai, an open-source visual discussion boa

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods

npx agentmods add instructions/chuck00lin/canvai/claude-md

Clone the repo

git clone --depth 1 https://github.com/chuck00lin/canvai

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/instructions/chuck00lin/canvai/claude-md.svg)](https://agentmods.dev/instructions/chuck00lin/canvai/claude-md)

Your own site

<a href="https://agentmods.dev/instructions/chuck00lin/canvai/claude-md"><img src="https://agentmods.dev/badge/instructions/chuck00lin/canvai/claude-md.svg" alt="Measured on agentmods" height="20"></a>

Per session 709 This file is loaded in full into every session.

When invoked 709 The same file — it is already loaded in full.

Security scan A 0 findings. Scan, not verified.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00709	$0.00709
Opus 5	$0.00354	$0.00354
Sonnet 5	$0.00142	$0.00142
Haiku 4.5	$0.00071	$0.00071

Measured 5d ago against content hash 03c30f544c88, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

canvai CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 25 lines

How it starts

The opening of the file, as written. The whole thing — 25 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md — canvai

Repo-native visual discussion boards for humans + AI agents. JSON Canvas files as source of truth, thin local server (watcher/WebSocket/MCP/ELK layout), React Flow web client, Obsidian as optional client.

Ground rules

This is a PUBLIC open-source repo. Internal discussion notes, meeting transcripts, experiments, and scratch tests never go here — they live in ../canvai-internal/ (outside this repo). When in doubt, put it there.
Public docs and code comments are English. README.zh-TW.md is the one translated file; keep it in sync on substantive README changes.
docs/design.md is the source of truth for design decisions (numbered D1–D5). Don't contradict it silently — change it (and say why) or open an issue.

Non-negotiable invariants (from the design doc)

.canvas round-trips preserve unknown fields (Advanced Canvas compatibility) and mirror each file's serialization dialect to keep git diffs minimal. Test-enforced: packages/canvas-kit/test/io.test.ts (byte-identical round-trips) and the hub e2e.
Agents never compute absolute coordinates — agent-reachable write paths go through semantic ops + ELK auto-layout; human-dragged nodes are pinned and auto-layout routes around them.
Mermaid is I/O only (embed in cards, one-way import/explode) — never the persistence format. Rationale: design doc D2.
Concurrency stays turn-based (atomic writes, last-write-wins, event feed) until Phase 2; no CRDT before then.

Status & layout

Current phase: Phase 0 + Phase 1 core shipped — packages/canvas-kit (round-trip io, projection, ops, diff, pinned-aware ELK layout), packages/hub (MCP stdio server + serve mode: watcher/WebSocket/REST), packages/web (React Flow client; Vite build, needs DOM lib so it has its own tsconfig and is excluded from the root one). The two hub processes coordinate purely through files (.canvas, .canvai/state.json, .canvai/events.jsonl). Next: field-test the live loop, measure the token bill of a real discussion turn, then Phase 2 (Yjs, presence, mermaid explode, @agent pins).
No build step for hub/kit: Node ≥ 23.6 runs the TypeScript directly (erasable syntax only — no enums/namespaces; relative imports use .ts extensions). Web builds with npm run web:build. npm test (vitest, includes MCP + serve e2e) and npm run typecheck must stay green.
Dev commands: npm run serve (HTTP/WS on 5199), npm run web:dev (Vite dev server proxying to 5199), npm run hub (MCP stdio).
Dogfooding: .mcp.json registers the hub for this repo; discuss/roadmap.canvas is a live board — feel free to update it via apply_ops when the roadmap changes.
License MIT. Do not vendor code from GPL projects (notably obsidian-advanced-canvas); format compatibility only.

Read the full file on GitHub · 25 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

5d ago First seen · 25 lines · 709 tokens per session scan A 03c30f544c88

Subscribe to this mod's changes

canvai CLAUDE.md is an instructions file published in the GitHub repository chuck00lin/canvai (2 stars, last pushed 1mo ago), licensed MIT. It adds 709 tokens to every session, about $0.0035 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

ai-agents canvas claude-code jsoncanvas mcp model-context-protocol

canvai CLAUDE.md

Getting it into your agent

Wrote this? Show the measurements

What it costs to keep this loaded

Grade A, and why

Nothing flagged

How it starts

CLAUDE.md — canvai

Ground rules

Non-negotiable invariants (from the design doc)

Status & layout

What this file has done since we first saw it

Other instructions, from other repositories

openrouter-mcp-multimodal AGENTS.md

llm-context.py CLAUDE.md

seekstone CLAUDE.md

Plonk AGENTS.md

cyxj-groksearch CLAUDE.md

mcp-dockhand copilot-review.instructions.md