Skip to content
MCP server

Figjam MCP

Local FigJam server that turns boards into cited context and tracks snapshot changes

Runs locallyStatus and history
Install
Pricing
Free
Platforms
self-hosted, CLI
Runs
Locally
Version
0.14.0
Listing updated
today

Similar MCP servers

About Figjam MCP

Figjam MCP is a free MCP server for productivity. It runs locally on your machine and is started by your AI client.

Workshop boards hold decisions and open questions that a screenshot flattens. This server reads a FigJam board through the Figma REST API, groups stickies, tables, images and arrows into clusters, and answers with links back to the original nodes, so a claim can be traced to its sticky.

Five tools cover the workflow: ingest_board saves a local snapshot, get_board_context pages through overviews and topic search, answer_from_board answers one specific question with node-level citations or says evidence is insufficient, diff_board compares snapshots including moved nodes, table cells and renamed sections, and diagnose_llm_config tests a configured model. Credentials found in a board are redacted before reaching the cache, output or a model.

A Figma personal access token with file_content:read is required, and each ingest is one Tier 1 file download, which the README warns is limited to a few per month on some plans. An OpenAI-compatible provider is optional and adds image descriptions and answers; without it, max_speed ingest works text-only. Node.js 22.12 or newer is needed.

Install

Run in a terminal

claude mcp add figjam-mcp --env FIGMA_ACCESS_TOKEN='<FIGMA_ACCESS_TOKEN>' --env LLM_API_KEY='<LLM_API_KEY>' -- npx -y figjam-context-mcp

Add to claude_desktop_config.json

{
  "mcpServers": {
    "figjam-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "figjam-context-mcp"
      ],
      "env": {
        "FIGMA_ACCESS_TOKEN": "<FIGMA_ACCESS_TOKEN>",
        "LLM_API_KEY": "<LLM_API_KEY>"
      }
    }
  }
}

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "figjam-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "figjam-context-mcp"
      ],
      "env": {
        "FIGMA_ACCESS_TOKEN": "<FIGMA_ACCESS_TOKEN>",
        "LLM_API_KEY": "<LLM_API_KEY>"
      }
    }
  }
}

Add to .vscode/mcp.json

{
  "servers": {
    "figjam-mcp": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "figjam-context-mcp"
      ],
      "env": {
        "FIGMA_ACCESS_TOKEN": "<FIGMA_ACCESS_TOKEN>",
        "LLM_API_KEY": "<LLM_API_KEY>"
      }
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "figjam-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "figjam-context-mcp"
      ],
      "env": {
        "FIGMA_ACCESS_TOKEN": "<FIGMA_ACCESS_TOKEN>",
        "LLM_API_KEY": "<LLM_API_KEY>"
      }
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "figjam-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "figjam-context-mcp"
      ],
      "env": {
        "FIGMA_ACCESS_TOKEN": "<FIGMA_ACCESS_TOKEN>",
        "LLM_API_KEY": "<LLM_API_KEY>"
      }
    }
  }
}

Using several servers? Build one config file with the MCP config generator.

Configuration

  • FIGMA_ACCESS_TOKENFigma personal access token with file_content:read; ingest_board reads the board with it.Required
  • LLM_API_KEYAPI key of the OpenAI-compatible provider in LLM_BASE_URL. Needed with LLM_BASE_URL for image interpretation and answers; a local server without sign-in needs any non-empty value.
  • LLM_BASE_URLAPI root of an OpenAI-compatible provider, for example https://openrouter.ai/api/v1. Needed with LLM_API_KEY for image interpretation and answers.
  • FIGJAM_MCP_CACHE_DIRPrivate local cache root; v4 snapshots are stored in a subdirectory.
  • FIGJAM_MCP_ENV_FILEAbsolute path of an env file to load; variables already set win. The server does not read a .env from its working directory.

Tools this server provides (5)

Taken from the project documentation; the running server may differ.

ingest_board
Read a board, cluster it and save a local snapshot
get_board_context
Read saved context with topic search and pagination
answer_from_board
Answer one question with node-level citations
diff_board
Compare snapshots of a board
diagnose_llm_config
Test the configured models with small checks

What it does

Clustered boards
Groups stickies, tables, images and arrows into sections.
Cited answers
Answers one question with node-level citations or an insufficient-evidence result.
Snapshot diffs
Reports moved nodes, changed cells, arrows and renamed sections.
Credential redaction
Removes explicit credentials before caching or model use.
Bring your own model
Uses any OpenAI-compatible provider for images and answers.

Good for

  • Ask what a workshop board decided
  • Track what changed on a board between sessions
  • Read a large board section by section
  • Pull table values from FigJam with a source link

Pricing

Free

No price listed; you provide your own Figma token and optional model key

Figjam MCP FAQ

What does it need from Figma?

A personal access token with the `file_content:read` scope.

Does it need an LLM?

Not for `max_speed` ingest; answers and image interpretation need an OpenAI-compatible provider.

Is there a rate limit?

Each ingest counts as a Tier 1 Figma request, and the README says some plans allow about 6 a month.

How does it differ from Figma's own MCP server?

It focuses on cited, trackable context and snapshot diffs; the README says the two work together.

Is Figjam MCP free?

Yes. Figjam MCP is free to use.

How do I add Figjam MCP to Claude Code?

Run this command in your terminal: claude mcp add figjam-mcp --env FIGMA_ACCESS_TOKEN='<FIGMA_ACCESS_TOKEN>' --env LLM_API_KEY='<LLM_API_KEY>' -- npx -y figjam-context-mcp

What are the best alternatives to Figjam MCP?

The closest alternatives to Figjam MCP are HyperFrames by HeyGen, next-ai-draw-io, Ha MCP, mcp-for-blender and ffmpeg-skill.

Where this comes from

  • Description: written from the project's README.
  • Install steps: from the official MCP Registry entry.

How AIPoweredBot checks servers