figma-comments-mcp
Local MCP server that lets Claude read, understand, and act on Figma comments — the layer the official Figma Dev Mode MCP doesn't expose.
Install / Use
claude mcp add junsimon -- npx -y github:junsimon/figma-comments-mcpIf the server publishes to npm under a different name, use that package instead — check the repo README.
MCP Server
Model Context Protocol server
Quality Score
Category
DesignSupported Platforms
Tags
Our assessment of figma-comments-mcp
figma-comments-mcp scores 78/100 on our quality scale, 236th of 260 Design skills we index.
Its MCP Server is 9.9 KB long, well organised into 18 sections with 10 code examples: a thorough specification that gives an agent plenty to work with.
It has 10 GitHub stars, so there is little community track record yet; judge it on its content.
Maintenance, license and trust
- The repository was last updated about 5 months ago. That is recent enough to be usable, but agent tooling moves fast, so check the instructions against your agent's current version.
- Our last check on 2026-11-11 found the source still online.
- It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 95/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
No issues foundOur scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.
AI review by kimi-k2.7-code on 2026-11-12. Automated pattern scan on 2026-11-12. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
figma-comments-mcp compared with similar skills
All 4 of these similar skills score higher than figma-comments-mcp; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| figma-comments-mcp (this skill)by junsimon | 78 | 10 | 5mo ago | MCP Server |
| Agent-Reachby Panniantong | 100 | 112.3k | 8d ago | CLAUDE.md |
| rufloby ruvnet | 100 | 81.9k | today | CLAUDE.md |
| headroomby headroomlabs-ai | 100 | 79.8k | today | CLAUDE.md |
| CowAgentby zhayujie | 100 | 48.0k | today | CLAUDE.md |
Frequently asked questions
- How do I install figma-comments-mcp?
- Run
claude mcp add junsimon -- npx -y github:junsimon/figma-comments-mcp. The install tabs above show the steps for each supported agent. - Which AI agents does figma-comments-mcp work with?
- It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
- Is figma-comments-mcp safe to use?
- Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is MIT-licensed and scores 95/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is figma-comments-mcp still maintained?
- The repository was last updated about 5 months ago. That is recent enough to be usable, but agent tooling moves fast, so check the instructions against your agent's current version.
Skill content
View source on GitHubfigma-comments-mcp
A local MCP server that lets Claude (or any MCP client) read, understand, and act on comments in Figma and FigJam — the layer the official Figma Dev Mode MCP doesn't expose.
Works on both Figma design files (figma.com/design/…) and FigJam boards (figma.com/board/…), so it fits the whole arc of a project: design reviews, design QA, and the messy collaborative middle — workshops, brainstorms, retros, and jams where the thinking lives in pinned comments on stickies and frames.
Why this exists: clients, reviewers, and workshop participants leave most of their feedback as comments pinned onto the canvas. AI assistants can read the design itself — layers, screenshots, code — but not those pinned comments, so the feedback never actually makes it into the work. This server closes that gap: Claude reads the comments and knows exactly which layer or sticky each one points to, so "make this blue" becomes "make Checkout / CTA Button blue."
The official Figma MCP gives an AI assistant the design tree, screenshots, variables, and code, but not comments. Clients, reviewers, and workshop participants leave most of their feedback as comments pinned onto the canvas. This server wraps the Figma REST comments API and, crucially, resolves which node each comment is pinned to — a layer in a design file, or a sticky/shape/section in a FigJam board (name + breadcrumb path + node id) — so feedback like "make this blue" becomes actionable: "comment on Checkout / CTA Button (node 1:360): make this blue."
It composes alongside the official Figma MCP — that one reads design structure and writes design changes; this one adds the missing comment layer.
┌─────────┐ reads comments ┌──────────────────────┐
│ Claude │ ───────────────────▶ │ figma-comments-mcp │ ──▶ Figma REST API
│ (MCP │ │ (this repo) │ /v1/files/:key/comments
│ client) │ ◀─────────────────── └──────────────────────┘
│ │ reads/writes design ┌──────────────────────┐
│ │ ◀──────────────────▶ │ Figma Dev Mode MCP │ ──▶ Figma (official)
└─────────┘ └──────────────────────┘
Contents
- Tools
- Sample output
- How comment → node mapping works
- Quick start
- Register with an MCP client
- CLI
- Configuration
- Development
- Contributing · Security · License
Tools
Read (always available):
| Tool | What it does |
|---|---|
| get_comments_summary | Counts (total/open/resolved), open threads grouped by target page/frame, one-line digest per open thread. Start here. |
| list_comments | All threads, grouped, each enriched with its target node. Open-only by default. |
| get_comment_thread | One full thread (root + replies + reactions + resolved state + target node). |
| render_comment_target | Renders the node a comment is pinned to (or an explicit node_id) and returns the image so the assistant can see it. |
Write (only when FIGMA_ENABLE_WRITES=true, and each defaults to a dry_run preview):
| Tool | What it does |
|---|---|
| reply_to_comment | Post a reply to a thread (e.g. "Done — changed the button to blue"). |
| react_to_comment | Add an emoji reaction (default :white_check_mark:) as a lightweight "addressed" marker. |
| post_comment | Place a new pinned comment at a node or coordinate. |
Resolution limitation: the Figma REST API has no endpoint to resolve a comment —
resolved_atcan only be set in the Figma UI. Use a reply or a:white_check_mark:reaction as the machine-writable "addressed" signal.
Sample output
get_comments_summary returns an at-a-glance overview — counts, open threads grouped by where they sit, and a one-line digest of each:
Comments: 12 total, 4 open, 8 resolved.
Open threads by target:
- Checkout / Payment: 2
- Home / Hero: 1
- (unpinned): 1
Open threads:
#3 Checkout / CTA Button: make this blue
#7 Checkout / Card field: the label is cut off on mobile
#9 Home / Hero: can we try a lighter headline?
#11 (unpinned): overall looks great, ship it
list_comments / get_comment_thread return the full thread, each tied to the exact layer it's pinned to:
#3 [OPEN] @jane 2026-06-10
Target: Checkout / CTA Button (node 1:360)
"make this blue"
└ reply @simon: "on it"
✅ x1
The (node 1:360) is the part the official Figma MCP can't give you: it tells the assistant precisely which layer the feedback is about.
How comment → node mapping works
Each comment carries client_meta in one of four shapes:
- FrameOffset / FrameOffsetRegion — has a
node_id→ direct lookup of the node name, type, and breadcrumb path (a layer in Figma, a sticky/shape/section in FigJam). - Vector / Region — bare absolute coordinates → the server fetches the file's node tree and returns the deepest node whose bounding box contains the point, flagged
[inferred]so you know the mapping is heuristic.
The file tree is fetched once and cached in-memory (TTL, default 60s) so enrichment doesn't refetch for every comment.
Quick start
Requirements: Node 20+ and npm.
1. Create a Figma Personal Access Token
Figma → Settings → Security → Personal access tokens. Select these scopes when creating it:
file_comments:read— read comments & reactions (required)file_content:read— read the node tree for enrichment + render images (required)file_comments:write— post replies/reactions/comments (only if you enable writes)
2. Install and build
git clone https://github.com/junsimon/figma-comments-mcp.git
cd figma-comments-mcp
npm install
npm run build
3. Provide the token
Copy .env.example to .env and set FIGMA_TOKEN:
cp .env.example .env
# edit .env and paste your token
The server and CLI auto-load .env from the project root. (Alternatively, pass
FIGMA_TOKEN via the MCP client's env block — see below.) .env is gitignored.
4. Verify with the CLI
node dist/cli.js "https://www.figma.com/design/<KEY>/<name>" --summary # Figma design file
node dist/cli.js "https://www.figma.com/board/<KEY>/<name>" --summary # FigJam board
You should see comment counts and a one-line digest per open thread. A bare file
key works too, and /design/, /board/, /file/, and /proto/ URLs are all accepted.
Register with an MCP client
The server speaks MCP over stdio. For Claude Code, either run:
claude mcp add figma-comments \
-e FIGMA_ENABLE_WRITES=false \
-- node /absolute/path/to/figma-comments-mcp/dist/index.js
…or add it to your MCP config (.mcp.json / claude_desktop_config.json / etc.):
{
"mcpServers": {
"figma-comments": {
"command": "node",
"args": ["/absolute/path/to/figma-comments-mcp/dist/index.js"],
"env": {
"FIGMA_TOKEN": "<your token>",
"FIGMA_ENABLE_WRITES": "false"
}
}
}
}
Replace
/absolute/path/to/figma-comments-mcpwith the real path on your machine — that string is a placeholder, not a literal value. If you copy it verbatim, the client will fail to start the server withError: Cannot find module '/absolute/path/to/figma-comments-mcp/dist/index.js'. Runpwdin the cloned repo to get the path, and make sure you've built it (npm run build) sodist/index.jsexists. The same applies to theclaude mcp addcommand above.
If you keep the token in .env, you can omit FIGMA_TOKEN from the config above.
Set FIGMA_ENABLE_WRITES to true to expose the write tools — even then, every
write defaults to a dry_run preview, so the assistant must pass dry_run:false to
actually mutate a shared file. Restart the client after registering.
CLI (for debugging)
node dist/cli.js "<figma file url>" # enriched comment blocks (open only)
node dist/cli.js "<figma file url>" --include-resolved # include resolved
node dist/cli.js "<figma file url>" --summary # counts + digest
node dist/cli.js "<figma file url>" --json # raw enriched JSON
During development you can skip the build with npm run cli -- "<url>" (via tsx).
Configuration reference
| Env var | Default | Purpose |
|---|---|---|
| FIGMA_TOKEN | — (required) | Personal Access Token. |
| FIGMA_ENABLE_WRITES | false | Register the write tools. |
| FIGMA_TREE_TTL_MS | 60000 | In-memory file-tree cache TTL. |
| FIGMA_TREE_DEPTH | 4 | Max node-tree depth fetched for pin enrichment. Bounds the payload so large files don't drop the connection. 0 = unbounded. |
All can be set in the environment or in .env.
Development
npm test # vitest — URL parsing, enrichment hit-test, threading
npm run typecheck # tsc --noEmit
npm run dev # run the server from source via tsx
npm run cli -- "<url>" # run the CLI from source
CI (GitHub Actions) runs typecheck, build, and tests on Node 20.
Contributing
Contributions are welcome — see CONTRIBUTING.md. Please don't commit secrets or paste tokens into issues/PRs.
Security
Keep your token in .env (gitignored), grant it the minimum scopes you need, and
revoke any token that leaks. The server only ever sends the token to Figma's API.
Found a vulnerability? Email simonjun@simonjun.cz rather than opening a public issue.
License
MIT © Šimon Jůn
Related Skills
Agent-Reach
112.3kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
ruflo
81.9k🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
headroom
79.8kCompress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
CowAgent
48.0kOpen-source personal AI assistant & Agent Harness. Plans tasks, runs tools and skills, self-evolves with memory and knowledge. Multi-agent, multi-model, multi-channel. Lightweight, extensible, one-line install.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
