Figma MCP bridge that connects Claude Code, Cursor, and other agents to a Figma plugin through WebSocket channels so they can read designs and modify nodes, text, layouts, styles, annotations, and components.
TalkToFigma can read and modify Figma documents, selected nodes, text content, annotations, auto-layout settings, fills, strokes, component instances, prototype reactions, and exported node images., Agents can create, move, resize, clone, and delete nodes, so use a duplicate file or branch-like design workflow before broad edits., Always join the intended WebSocket channel and verify the active Figma document before running write tools., Review batch text replacements, component override propagation, annotation conversion, and connector generation before sharing or handing off design files., Keep the WebSocket bridge local or otherwise protected so untrusted clients cannot attach to an active Figma session.
Privacy notes
Figma document structure, node names, selected content, annotations, component metadata, local styles, exported images, text content, and prototype information may be exposed to the MCP client and model., Design files often contain unreleased product UI, customer data in mockups, brand assets, internal names, tokens, pricing, screenshots, or partner/client IP., Exported images can include visual details that are more sensitive than node metadata alone., Channel names, WebSocket messages, local logs, and MCP prompts may reveal active file context and design intent.
Author
Grab
Submitted by
oktofeesh1
Claim status
unclaimed
Last verified
2026-06-06
Decision playbook
Review trust signals before you adopt
Signals are present but mixed. Use the checklist below to confirm the source and operational safety for your environment.
Compare context
Selected
0
Current score
63
Baseline
—
Delta
No baseline selected
No major trust-signal divergence detected in the current selection.
Source and provenance checks
Needs review
Confirm ownership and provenance before trusting install instructions.
Source link availableRequired
Open the canonical repository and verify ownership.
Done
Source provenance statusRequired
Marked as source-backed.
Done
Metadata reviewed
No reviewed flag detected in metadata.
Pending
Safety and privacy checks
Complete
Validate risk disclosures before installation or API wiring.
Safety notes presentRequired
Review the listed safety guidance before running commands.
Done
Privacy notes presentRequired
Review data handling notes before connecting accounts or secrets.
Done
Trust level risk gateRequired
Trust level does not block evaluation.
Done
Package and install checks
Needs review
Check package metadata and artifact integrity signals.
Install payload available
Install or copy payload is available for review.
Done
Package verification flag
No package verification flag provided.
Pending
Checksum metadata
No checksum provided for downloaded artifact.
Pending
Compare-driven decision checks
Needs review
Use compare context to validate trade-offs before adoption.
Compare tray has multiple entries
Add at least one more entry to compare trust differences.
5 safety and 4 privacy notes across 3 risk areas. Review closely: credentials & tokens.
3 areas
SafetyGeneralTalkToFigma can read and modify Figma documents, selected nodes, text content, annotations, auto-layout settings, fills, strokes, component instances, prototype reactions, and exported node images.
SafetyLocal filesAgents can create, move, resize, clone, and delete nodes, so use a duplicate file or branch-like design workflow before broad edits.
SafetyLocal filesAlways join the intended WebSocket channel and verify the active Figma document before running write tools.
SafetyLocal filesReview batch text replacements, component override propagation, annotation conversion, and connector generation before sharing or handing off design files.
SafetyCredentials & tokensKeep the WebSocket bridge local or otherwise protected so untrusted clients cannot attach to an active Figma session.
PrivacyGeneralFigma document structure, node names, selected content, annotations, component metadata, local styles, exported images, text content, and prototype information may be exposed to the MCP client and model.
PrivacyCredentials & tokensDesign files often contain unreleased product UI, customer data in mockups, brand assets, internal names, tokens, pricing, screenshots, or partner/client IP.
PrivacyGeneralExported images can include visual details that are more sensitive than node metadata alone.
PrivacyLocal filesChannel names, WebSocket messages, local logs, and MCP prompts may reveal active file context and design intent.
Safety notes
TalkToFigma can read and modify Figma documents, selected nodes, text content, annotations, auto-layout settings, fills, strokes, component instances, prototype reactions, and exported node images.
Agents can create, move, resize, clone, and delete nodes, so use a duplicate file or branch-like design workflow before broad edits.
Always join the intended WebSocket channel and verify the active Figma document before running write tools.
Review batch text replacements, component override propagation, annotation conversion, and connector generation before sharing or handing off design files.
Keep the WebSocket bridge local or otherwise protected so untrusted clients cannot attach to an active Figma session.
Privacy notes
Figma document structure, node names, selected content, annotations, component metadata, local styles, exported images, text content, and prototype information may be exposed to the MCP client and model.
Design files often contain unreleased product UI, customer data in mockups, brand assets, internal names, tokens, pricing, screenshots, or partner/client IP.
Exported images can include visual details that are more sensitive than node metadata alone.
Channel names, WebSocket messages, local logs, and MCP prompts may reveal active file context and design intent.
Prerequisites
Bun installed through a trusted installation path.
Figma desktop or browser session with access to the design file you are authorized to inspect or modify.
The TalkToFigma Figma plugin installed from the documented upstream path or linked locally from the repository.
WebSocket bridge running with `bun socket`.
MCP client such as Claude Code, Cursor, or another compatible host configured with the `cursor-talk-to-figma-mcp` package.
Schema details
Install type
cli
Troubleshooting
No
Source repository stats
Scope
Source repo
Collection metadata
Estimated setup
20 minutes
Difficulty
advanced
Full copyable content
bunx cursor-talk-to-figma-mcp@latest
About this resource
Content
TalkToFigma MCP Server connects an MCP client to Figma through a TypeScript MCP
server, a local WebSocket bridge, and a Figma plugin. It lets agents read Figma
documents and selected nodes, then perform design operations such as creating
frames and text, changing layout settings, editing styles, replacing copy,
managing annotations, exporting nodes, and working with component instances.
This is a write-capable design automation bridge. It is distinct from
read-only Figma context extraction because it can modify the active Figma file
through the plugin once the WebSocket channel is connected.
These sources were reviewed on 2026-06-06. Prefer the live repository,
README, package metadata, and npm registry metadata for current commands,
server package names, WebSocket setup, plugin setup, and supported Figma tools.
Features
Read Figma document and selection metadata.
Inspect individual or multiple node records.
Create rectangles, frames, text nodes, component instances, and connectors.
Start the WebSocket bridge from a checkout of the repository:
bun socket
Open Figma, run the TalkToFigma plugin, join a channel with join_channel, and
verify the active file before allowing write tools.
Use Cases
Let Claude inspect selected Figma nodes while implementing UI.
Create rough frames, text, rectangles, and component instances from prompts.
Batch update design copy across many text nodes.
Convert manual annotations into native Figma annotations.
Propagate component instance overrides to repeated UI patterns.
Generate connector lines from prototype reactions for FigJam-style flow maps.
Safety and Privacy
Start with read tools such as get_document_info, get_selection, and
read_my_design. Move to write tools only after confirming the active Figma
document, selected node, and WebSocket channel.
Use duplicate files for experiments. Batch operations such as text replacement,
annotation conversion, component override propagation, and node deletion can
change large parts of a design quickly.
Treat design data as sensitive model context. Node names, text layers,
component metadata, exported images, and prototype flows can reveal product
strategy, unreleased UI, customer information, brand assets, and implementation
details.
Duplicate Check
No grab/cursor-talk-to-figma-mcp entry, TalkToFigma entry, or matching
TalkToFigma source URL was found in content/mcp. Existing Figma-related
entries cover different projects and do not document this write-capable
TalkToFigma bridge.
Show that TalkToFigma MCP Server is listed on HeyClaude. Paste this Markdown into your README — it renders the badge and links back to this page.
[](https://heyclau.de/entry/mcp/talktofigma-mcp-server)
How it compares
TalkToFigma MCP Server side by side with 3 alternatives on trust, install, platform support, and disclosed safety notes — all from reviewed registry metadata.
Figma MCP bridge that connects Claude Code, Cursor, and other agents to a Figma plugin through WebSocket channels so they can read designs and modify nodes, text, layouts, styles, annotations, and components.
MCP server for extracting design systems from live websites, including design tokens, regions, components, contrast data, Tailwind themes, Figma variables, and prompt packs.
Official MCP server for agent-device, Callstack's device automation CLI for inspecting, controlling, debugging, recording, and collecting evidence from iOS, Android, TV, macOS, Linux, React Native, Expo, Flutter, and native apps.
✓TalkToFigma can read and modify Figma documents, selected nodes, text content, annotations, auto-layout settings, fills, strokes, component instances, prototype reactions, and exported node images.
Agents can create, move, resize, clone, and delete nodes, so use a duplicate file or branch-like design workflow before broad edits.
Always join the intended WebSocket channel and verify the active Figma document before running write tools.
Review batch text replacements, component override propagation, annotation conversion, and connector generation before sharing or handing off design files.
Keep the WebSocket bridge local or otherwise protected so untrusted clients cannot attach to an active Figma session.
✓designlang uses Playwright to crawl live pages and can capture DOM-derived styles, responsive behavior, interaction states, screenshots, and accessibility findings.
Authenticated extraction options can use cookies, cookie files, headers, and custom user agents, so never pass production session cookies or credentials unless approved.
Generated outputs may include design tokens, Tailwind config, shadcn variables, Figma variables, component anatomy, prompts, screenshots, reports, and cloned starter code.
Commands such as apply, clone, sync, drift, visual-diff, and MCP extraction can read live websites and write files in the configured output or project directory.
Review extracted prompt packs and generated code before using them in another agent workflow or committing them.
✓Agent Device MCP exposes structured tools backed by `AgentDeviceClient`; the docs state it does not expose generic shell execution over MCP.
Tools and CLI workflows can open apps, inspect UI, tap, type, scroll, perform gestures, wait, assert state, handle alerts, and close sessions.
Evidence workflows can capture screenshots, recordings, logs, traces, network traffic, performance samples, crash context, React profiles, and replay files.
Mutating commands should run serially against one session, and separate sessions or devices should be used for parallel work.
Prefer dedicated test devices or simulators, and require approval before entering credentials, submitting forms, changing settings, installing apps, sending messages, or touching production accounts.
✓AgentQL MCP exposes one tool, `extract-web-data`, that sends a target URL and natural-language extraction prompt to the AgentQL API.
The tool is intended for public webpages; do not use it to bypass access controls, scrape private pages, evade paywalls, or extract data where automated collection is prohibited.
Web extraction can still trigger target-site rate limits, legal restrictions, robots guidance, or terms-of-service concerns.
The source implementation uses AgentQL's query-data endpoint with fast mode, no screenshot capture, no scroll-to-bottom behavior, and no local browser cookies.
Treat extracted output as untrusted web data that may include errors, stale content, ads, tracking text, or prompt-injection attempts.
Privacy notes
✓Figma document structure, node names, selected content, annotations, component metadata, local styles, exported images, text content, and prototype information may be exposed to the MCP client and model.
Design files often contain unreleased product UI, customer data in mockups, brand assets, internal names, tokens, pricing, screenshots, or partner/client IP.
Exported images can include visual details that are more sensitive than node metadata alone.
Channel names, WebSocket messages, local logs, and MCP prompts may reveal active file context and design intent.
✓URLs, page content, DOM text, CSS, screenshots, fonts, images, design tokens, cookies, headers, prompts, tool arguments, reports, and generated files may be visible to the MCP client and model provider.
Authenticated or internal sites can expose product plans, unreleased UI, customer data, analytics identifiers, private brand assets, and implementation details.
Output directories can retain extracted website data after the MCP session ends.
Avoid running the server against private, paid, internal, or authenticated properties without legal and security approval.
✓Screenshots, recordings, traces, logs, network dumps, replay files, reports, UI snapshots, typed input, and React profiles can contain private UI state, tokens, request data, customer information, or credentials.
macOS, iOS, Android, and TV automation can expose local app state, notifications, device names, package identifiers, app content, system dialogs, and permission prompts.
Network inspection artifacts may include headers, payloads, session identifiers, URLs, and API data; review before sharing or committing.
Interactive CLI runs may check npm for newer package versions unless `AGENT_DEVICE_NO_UPDATE_NOTIFIER=1` is set.
✓Target URLs, extraction prompts, API key-authenticated requests, and extracted structured data are sent to AgentQL's API.
Extracted data can include personal data, copyrighted content, customer information, job postings, prices, social content, or other third-party material.
AGENTQL_API_KEY should stay out of prompts, issues, logs, screenshots, and committed configuration files.
Claude transcripts and downstream reports may retain extracted data, so avoid collecting information that is not approved for the model session.
Prerequisites
Bun installed through a trusted installation path.
Figma desktop or browser session with access to the design file you are authorized to inspect or modify.
The TalkToFigma Figma plugin installed from the documented upstream path or linked locally from the repository.
WebSocket bridge running with `bun socket`.
Node.js 20 or newer available to the MCP client runtime.
Network access to the websites you plan to extract.
Permission to crawl and analyze the target sites.
Playwright or Chromium installation behavior reviewed for the local environment.
Node.js 22 or newer and `agent-device` installed globally or project-locally.
Xcode tooling for iOS, tvOS, or macOS targets, or Android SDK and ADB for Android targets.
Device, simulator, emulator, TV, macOS, or Linux desktop target that the agent is allowed to automate.
Required local permissions such as Android device trust, iOS Developer Mode, macOS Accessibility, and Screen Recording where applicable.
Node.js and npx available to the MCP client runtime.
AgentQL API key from the AgentQL developer portal.
Approved list of public webpages or domains Claude may query.
Review of target site terms, robots guidance, rate limits, and data-use rules before extraction.
Install
bunx cursor-talk-to-figma-mcp@latest
npx -y designlang mcp
claude mcp add --transport stdio --scope user agent-device -- agent-device mcp