High-performance code intelligence MCP server. Indexes codebases into a persistent knowledge graph — average repo in milliseconds. 158 languages, sub-ms queries, 99% fewer tokens. Single static binary, zero dependencies.
Figma plugin and MCP server for agent access
This repo connects Figma to an MCP client through a browser plugin and a local server. The server keeps WebSocket connections to one or more open Figma files, then routes tool calls by `fileKey` so an agent can inspect or edit the right file. It also exposes safe write tools like text edits, node changes, screenshots, and basic layout actions.
Builders who want their agent to read Figma documents, inspect selections, and make limited edits from an MCP client.
You can work on Figma-driven design tasks with an agent without re-explaining the file or burning through API request limits.
What it does
Multi-file Figma connections
Keeps multiple Figma files connected at once and routes requests by `fileKey`.
Document and node inspection
Provides tools like `get_document`, `get_selection`, `get_node`, `get_styles`, `get_metadata`, and `get_design_context`.
Design token access
Reads variable collections, modes, and values with `get_variable_defs`.
Export and screenshot tools
Exports nodes as PNG, SVG, JPG, or PDF, and can save screenshots to the local filesystem.
Safe write tools
Lets an agent edit text, node properties, fills, effects, strokes, auto-layout, visibility, and basic structure.
Basic creation and import
Creates pages, frames, text, shapes, and images, and can import HTML layer trees from `html-figma` JSON.
Animation tools
Includes beta tools for reading and applying motion styles and keyframe tracks.
Leader and follower server setup
Uses leader election so only one MCP server keeps the active WebSocket connection while others proxy through HTTP.
How to get it
- 1Install the root dependencies once. This runs Husky's prepare script, which installs the…
cd figma-mcp-bridge && bun install
README
Figma MCP Bridge
While other amazing Figma MCP servers like Figma-Context-MCP exist, one issues is the API limiting for free users.
The limit for free accounts is 6 requests per month, yes per month.
Figma MCP Bridge is a solution to this problem. It is a plugin + MCP server that streams live Figma document data to AI tools without hitting Figma API rate limits, so its Figma MCP for the rest of us ✊
It supports multiple Figma files connected simultaneously; open the plugin in each file and your AI agent can query any of them by fileKey. Single-file setups work exactly as before with no changes required.
It also includes a small, opt-in set of write tools for safe agent-driven edits — see Editing Notes below.
Demo
Watch a demo of building a UI in Cursor with Figma MCP Bridge
Quick Start
1. Add the MCP server to your favourite AI tool
Add the following to your AI tool's MCP configuration (e.g. Cursor, Windsurf, Claude Desktop):
{
"figma-bridge": {
"command": "npx",
"args": ["-y", "@gethopp/figma-mcp-bridge"]
}
}
That's it — no binaries to download or install.
2. Add the Figma plugin
Download the plugin from the latest release page, then in Figma go to Plugins > Development > Import plugin from manifest and select the manifest.json file from the plugin/ folder.
3. Start using it 🎉
Open a Figma file, run the plugin, and start prompting your AI tool. The MCP server will automatically connect to the plugin.
To work across multiple files, just open the plugin in each Figma file. The bridge keeps all connections active and your AI agent can target any of them by fileKey.
If you want to know more about how it works, read the How it works section.
Available Tools
| Tool | Description |
|---|---|
list_files | List all connected Figma files (supports multi-file workflows) |
get_document | Get the current Figma page document tree |
get_selection | Get the currently selected nodes in Figma |
get_node | Get a specific Figma node by ID (colon format, e.g. 4029:12345) |
get_styles | Get all local paint, text, effect, and grid styles |
get_metadata | Get file name, pages, and current page info |
get_design_context | Get a depth-limited tree optimized for understanding design context |
get_variable_defs | Get all variable collections, modes, and values (design tokens) |
get_screenshot | Export nodes as PNG/SVG/JPG/PDF (base64-encoded) |
save_screenshots | Export and save screenshots directly to the local filesystem |
get_motion_styles | List all available animation presets (beta) |
get_node_motion | Read a node's current animation styles and properties (beta) |
apply_animation_style | Apply a preset animation style to a node (beta) |
remove_animation_style | Remove an applied animation style from a node (beta) |
apply_manual_keyframe_track | Apply a manual keyframe track to a node property (beta) |
remove_manual_keyframe_track | Remove a manual keyframe track from a node property (beta) |
set_timeline_duration | Set the duration of a timeline in seconds (beta) |
set_node_visibility | Show or hide specific nodes |
set_text_content | Replace the contents of a text node |
set_text_properties | Patch font, size, alignment, auto-resize, color, and bounds on a text node |
set_node_properties | Patch common node properties: name, position, size, visibility, opacity, corner radius |
set_solid_fill | Replace a node's fill or stroke with a single solid paint |
set_gradient_fill | Replace a node's fill or stroke with a linear/radial/angular/diamond gradient |
set_effects | Replace a node's effects list (drop/inner shadows, layer/background blurs) |
set_stroke_properties | Patch stroke weight, align, dash pattern, cap, and join |
set_auto_layout | Configure auto-layout direction, padding, gap, alignment, sizing, and wrap |
create_page | Create a new page in the document, optionally switching to it |
create_frame | Create a new frame, optionally under a parent |
create_text | Create a new text node |
create_shape | Create a rectangle, ellipse, or line |
create_image | Create an image-backed rectangle from a local path, URL, or data URI |
import_html_layers | Bulk-import an html-figma layer tree (JSON) as frames, text, rectangles, and vectors |
duplicate_nodes | Duplicate nodes in place |
reparent_nodes | Move nodes into another parent |
group_nodes | Wrap a list of nodes (sharing a parent) in a new group |
ungroup_node | Ungroup a group or frame — children move up to its parent |
set_selection | Set the page selection to a list of node IDs (works in Dev Mode) |
scroll_and_zoom_into_view | Frame the viewport around the given nodes (works in Dev Mode) |
delete_nodes | Delete nodes with explicit confirmation |
All tools accept an optional fileKey parameter when multiple Figma files are connected. Use list_files to discover connected files and their keys.
Editing Notes
- Edit tools work only when the plugin is opened in Figma's design editor (Dev Mode is read-only — they will return a clear error there).
- The current user must have permission to edit the target file.
delete_nodesis intentionally gated behindconfirm: true.- Text edits automatically load the fonts currently used by the target text node before applying the new content.
- New text nodes default to
Inter Regularunless a font is provided. create_imagereads local paths relative to the MCP server working directory unless you pass an absolute path.import_html_layerstakes a JSON file produced by html-figma's browserhtmlToFigma(). The path resolves relative to the MCP server working directory and must stay inside it, even when absolute. Everything lands inside one wrapper frame, and the response reportslayerCountagainstexpectedLayerCountso partial imports are visible.create_pagereturns the new page's ID — pass it asparentIdtocreate_frame/create_text/create_shape/create_imageto author content on that page without switching the editor.
What You Can Build
With the current write surface, an agent can build a basic slide deck in a new empty Figma file: create slide frames, style titles and body copy, lay out rectangles/ellipses/lines for cards and dividers, duplicate slide templates, reparent content into the right frame, and adjust common geometry/visual properties — including solid/gradient paints, shadows and blurs, stroke geometry, and auto-layout configuration.
The current version is intentionally limited — no components/instances, no variables/styles authoring, no per-segment text styling, and no vector boolean operations yet.
Local development
This repo uses Bun as its package manager and script runner throughout. Install it first if you don't have it.
1. Clone this repository locally
git clone git@github.com:gethopp/figma-mcp-bridge.git
2. Install root tooling
Install the root dependencies once. This runs Husky's prepare script, which installs the Git pre-commit hook that formats staged files with Prettier.
cd figma-mcp-bridge && bun install
3. Build the server
cd server && bun install && bun run build
4. Build the plugin
cd plugin && bun install && bun run build
5. Add the MCP server to your favourite AI tool
For local development, add the following to your AI tool's MCP config:
{
"figma-bridge": {
"command": "node",
"args": ["/path/to/figma-mcp-bridge/server/dist/index.js"]
}
}
Code style
The repo is formatted with Prettier (config in .prettierrc). A Husky pre-commit hook runs lint-staged, which formats only your staged files, so commits stay formatted automatically. You can also run it manually:
bun run format # format the whole repo
bun run format:check # verify formatting without writing (useful in CI)
Structure
Figma-MCP-Bridge/
├── plugin/ # Figma plugin (TypeScript/React)
└── server/ # MCP server (TypeScript/Node.js)
└── src/
├── index.ts # Entry point
├── bridge.ts # WebSocket bridge to Figma plugin
├── leader.ts # Leader: HTTP server + bridge
├── follower.ts # Follower: proxies to leader via HTTP
├── node.ts # Dynamic leader/follower role switching
├── election.ts # Leader election & health monitoring
├── tools.ts # MCP tool definitions
└── types.ts # Shared types
How it works
There are two main components to the Figma MCP Bridge:
1. The Figma Plugin
The Figma plugin is the user interface for the Figma MCP Bridge. You run this inside the Figma file you want to use the MCP server for, and its responsible for getting you all the information you need.
2. The MCP Server
The MCP server is the core of the Figma MCP Bridge. It maintains a registry of WebSocket connections keyed by fileKey, so multiple Figma files can be connected simultaneously. The server is responsible for:
- Handling WebSocket connections from one or more Figma plugin instances
- Routing tool calls to the correct file based on
fileKey - Forwarding responses back to the AI client
- Handling leader election (as we can have only one WS connection to an MCP server at a time)
┌─────────────────────────────────────────────────────────────────────────────┐
│ FIGMA (Browser) │
│ ┌───────────────────────────────────────────────────────────────────────┐ │
│ │ Figma Plugin │ │
│ │ (TypeScript/React) │ │
│ └───────────────────────────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────────────────────┘
│
│ WebSocket
│ (ws://localhost:1994/ws)
▼
┌─────────────────────────────────────────────────────────────────────────────┐
│ PRIMARY MCP SERVER │
│ (Leader on :1994) │
│ ┌─────────────────────────────────────────────────────────────────────┐ │
│ │ Bridge Endpoints: │ │
│ │ • Manages WebSocket conn • /ws (plugin) │ │
│ │ • Forwards requests to plugin • /ping (health) │ │
│ │ • Routes responses back • /rpc (followers) │ │
│ └─────────────────────────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────────────────────┘
▲ ▲
│ HTTP /rpc │ HTTP /rpc
│ POST requests │ POST requests
│ │
┌─────────────────┴───────────┐ ┌─────────────┴───────────────┐
│ FOLLOWER MCP SERVER 1 │ │ FOLLOWER MCP SERVER 2 │
│ │ │ │
│ • Pings leader /ping │ │ • Pings leader /ping │
│ • Forwards tool calls │ │ • Forwards tool calls │
│ via HTTP /rpc │ │ via HTTP /rpc │
│ • If leader dies → │ │ • If leader dies → │
│ attempts takeover │ │ attempts takeover │
└─────────────────────────────┘ └─────────────────────────────┘
▲ ▲
│ │
│ MCP Protocol │ MCP Protocol
│ (stdio) │ (stdio)
▼ ▼
┌─────────────────────────────┐ ┌─────────────────────────────┐
│ AI Tool / IDE 1 │ │ AI Tool / IDE 2 │
│ (e.g., Cursor) │ │ (e.g., Cursor) │
└─────────────────────────────┘ └─────────────────────────────┘
Files in the repo
- .github
- .husky
- plugin
- server
- .gitignore
- .nvmrc
- .prettierignore
- .prettierrc
- bun.lock
- LICENSE.md
- logo.png
- package.json
- README.md
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More connectors

Universal provider proxy for OpenAI Codex & Claude Code — use any LLM (Claude, Gemini, Grok, DeepSeek, Ollama…) with Codex CLI, App, SDK, and Claude Code
Git-native persistent memory for AI coding agents. Implements Google OKF v0.2 with sub-300µs in-memory BM25 search, embedded MCP server, and progressive disclosure. Slashes token bloat by 80% with zero external databases or dependencies. Built in pure Go.
Local-first code intelligence graph for MCP and CLI. Builds a persistent map of your codebase so AI coding tools read only what matters, with benchmarked context reductions on reviews and large-repo workflows.
Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is the proving ground. MCP server + CLI.
20 MB lightweight cross-platform database client for 90+ databases, including MySQL, PostgreSQL, SQLite, Redis, MongoDB, DuckDB, SQL Server, and Dameng. Built-in AI, MCP Server, CLI, desktop and Docker. | 轻量级跨平台数据库管理工具,支持 MySQL、PostgreSQL、SQLite、Redis、MongoDB、达梦等 90+ 数据库,提供桌面端、Docker、CLI、内置 AI 助手和 MCP Server。
