Integration
Use Figma with Claude, GPT, Gemini and more
Connect Figma once, then ask any model in llmwise to read from it and act in it: 52 actions, with every write shown to you before it runs.
Actions (toolkit version 20260721_00) checked against Composio's Figma toolkit page.
Short answer
Figma connects to llmwise in one click, and every model can then use its 52 actions: 41 that read and run on their own, and 11 that write and ask you first (7 of them every time). On paid plans, included in the message price.
Try it: Figma in one message
Send this with any model. The model finds Figma's actions for the job and runs them, and the reply shows each one.
Summarize the open comments on the Checkout redesign file in Figma, then reply with a comment that lists the three we'll fix first.
| Action | In llmwise |
|---|---|
| Get comments in a file | Reads: runs without asking |
| Add a comment to a file | Always asks first |
The reads run straight away; the write waits on an approval card with Allow and Deny, and shows what it will do first.
What the model can do in Figma
52 actions in all. Below: the 14 Composio marks as Figma's main ones, and 16 more. The other 22 are there too, and the model finds them when a job needs one.
Reads: runs without asking
- Design tokens to tailwind
- Download Figma Images
- Extract design tokens
- Extract Prototype Interactions
- Get component
- Get current user
- Get file component sets
- Get file components
- Get file json
- Get file metadata
- Get file nodes
- Get files in a project
- Get local variables
- Get projects in a team
- Detect Background Layers
- Discover Figma Resources
- Get activity logs
- Get comments in a file
- Get component set
- Get dev resources
- Get file styles
- Get image fills
- Get library analytics component action data
Writes: asks first
- Create dev resources
Always asks first
- Add a comment to a file
- Add a reaction to a comment
- Create, modify, or delete variables
- Delete a comment
- Delete a reaction
- Delete dev resource
What asks first, and why
Reading needs no approval. Anything that writes asks first, and sending, deleting, payments and sharing always ask.
In Figma, that means add a comment to a file, add a reaction to a comment, create, modify, or delete variables, and delete a comment always wait for your yes.
Always allow can switch off the question for one of Figma's 4 other writes, and a reply that has read outside content asks again anyway. One reply runs up to 10 actions, each result read up to 50,000 characters.
Where your Figma data goes
Each action runs through Composio and Figma's own service, which receive its inputs; the result is shown in your chat and saved with it. Composio holds the sign-in, encrypted. More in the Privacy Policy.
More apps
100+ apps connect in one click · 1,500+ in Composio's catalog. Every app with its own page, or add a remote server of your own with the MCP client.
Questions
How do I connect Figma to llmwise?
In Settings → Connectors, press Connect next to Figma and sign in on Figma's own page: one click, no setup. Composio holds the connection, and llmwise never sees your password. On a paid plan.
Which AI models work with Figma, and what does it cost?
Any of the 15 models in llmwise, from GLM 5.3 Flash to Claude Fable 5.1: Figma's actions are the same whichever you pick. Connectors are on the paid plans (Pro ($20 a month), Max ($50 a month), Ultra ($100 a month), and Studio ($200 a month)), included in the message price.
Claude, GPT, Gemini, DeepSeek, Grok, Kimi, and GLM, in one chat.
See what a message costs before you send it. Free is 5 messages to try; sign in with an email link, no password or card.