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.
npx agentmods add instructions/grab/cursor-talk-to-figma-mcp/agents-mdgit clone --depth 1 https://github.com/grab/cursor-talk-to-figma-mcpWhat 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 | $0.00911 | $0.00911 |
| Opus 5 | $0.00456 | $0.00456 |
| Sonnet 5 | $0.00182 | $0.00182 |
| Haiku 4.5 | $0.00091 | $0.00091 |
Grade A, and why
cursor-talk-to-figma-mcp AGENTS.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 yesterday.
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.
Copies of this mod
7 near-identical copies found in the catalogue:
- figma-local-bridge AGENTS.md — 100% identical, 0 lines differ
- talk-to-figma-mcp AGENTS.md — 100% identical, 0 lines differ
- edgen-figma-to-code-mcp AGENTS.md — 100% identical, 0 lines differ
- cursor-talk-to-figma-mcp CLAUDE.md — 91% identical, 21 lines differ
- edgen-figma-to-code-mcp CLAUDE.md — 91% identical, 21 lines differ
- figma-local-bridge CLAUDE.md — 91% identical, 21 lines differ
- talk-to-figma-mcp CLAUDE.md — 91% identical, 21 lines differ
How it starts
The opening of the file, as written. The whole thing — 80 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
This file provides guidance to Claude Code agents (claude.ai/code) when working with code in this repository.
Project Overview
MCP (Model Context Protocol) server that bridges AI agents (Cursor, Claude Code) with Figma. Three components communicate in a pipeline:
Claude Code / Cursor ←(stdio)→ MCP Server ←(WebSocket)→ WebSocket Relay ←(WebSocket)→ Figma Plugin
Build & Development Commands
bun install # Install dependencies
bun run build # Build MCP server (tsup → dist/)
bun run dev # Build in watch mode
bun socket # Start WebSocket relay server (port 3055)
bun run start # Run built MCP server
bun setup # Full setup (install + write .cursor/mcp.json + .mcp.json)
There is no test suite or linter configured.
Architecture
MCP Server (src/talk_to_figma_mcp/server.ts)
The main server implementing the MCP protocol via @modelcontextprotocol/sdk. Exposes 50+ tools (create shapes, modify text, manage layouts, export images, etc.) and several AI prompts (design strategies). Communicates with the AI agent over stdio and with the WebSocket relay via ws. Each request gets a UUID, is tracked in a pendingRequests Map with timeout/promise callbacks, and resolves when the plugin responds.
WebSocket Relay (src/socket.ts)
Lightweight Bun WebSocket server on port 3055 (configurable via PORT env). Routes messages between MCP server and Figma plugin using channel-based isolation. Clients call join to enter a channel; messages broadcast only within the same channel.
Figma Plugin (src/cursor_mcp_plugin/)
Runs inside Figma. code.js is the plugin main thread handling 30+ commands via a dispatcher. ui.html is the plugin UI for WebSocket connection management. manifest.json declares permissions (dynamic-page access, localhost network). The plugin is not built/bundled — code.js is written directly as the runtime artifact.
Key Patterns
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.
- yesterday First seen · 80 lines · 911 tokens per session scan A 505d9248890b
cursor-talk-to-figma-mcp AGENTS.md is an instructions file published in the GitHub repository grab/cursor-talk-to-figma-mcp (6,989 stars, last pushed 1mo ago), licensed MIT. It adds 911 tokens to every session, about $0.0046 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 2026-08-30.
Other instructions, from other repositories
openagent CLAUDE.md
Claude Code instructions for the-open-agent/openagent, covering claude.md, commands, architecture, backend (go / beego) and frontend (react).
pi-agent-dashboard AGENTS.md
Instructions for BlackBeltTechnology/pi-agent-dashboard, covering pi dashboard, docs-first gate — kb before grep (per-turn doctrine), code instructions (per-turn doctrine), investigation protocol — index first and documentation update protocol.
agentscope-java code-review.instructions.md
Instructions for agentscope-ai/agentscope-java, covering java code standards, security check, performance check, documentation sync check and pr title check.
hands-on-modern-rl AGENTS.md
AGENTS.md instructions for walkinglabs/hands-on-modern-rl, covering repository guidelines, project structure, commands, editing rules and tikz → svg workflow.
elephant-agent AGENTS.md
AGENTS.md instructions for agentic-in/elephant-agent, covering elephant agent entry, product north star, read first, non-negotiable rules and canonical commands.
elephant-agent copilot-instructions.md
Copilot instructions for agentic-in/elephant-agent, covering github copilot instructions, source of truth order, review priorities, harness-specific checks and validation expectations.