Skip to main content

Your coding agent,
on the canvas.

Claude Code, Codex, Cursor — any agent that speaks MCP reads your design as data and builds it in your codebase.

A designer working at a studio desk, a colleague behind them
Your agent, reading the canvas

Bring the agent
you already use.

Lindi runs an MCP server for every file. Any coding agent that can connect to one over HTTP can read the design.

  • Claude Code.

    Add Lindi as an MCP server with one command, then ask Claude Code to build the screen you are looking at.

  • Codex.

    Point Codex at the same endpoint with the same token. It fetches the design context and writes the component.

  • Cursor, and the rest.

    Cursor and any other tool that speaks MCP connect the same way — one endpoint, one token per file.

One command.
Then it reads the design.

Create a token for the file in Dev Mode and copy the endpoint beside it. Add both to your agent, and it can ask for a screen, its components, its variables — and everything needed to build it.

  • Read-only: it reads the design, it cannot change it
  • One token per file — revoke it and the connection ends
  • The same context Dev Mode shows a person
claude mcp add --transport http lindi \  https://lindi.ai/api/mcp \  --header "Authorization: Bearer lmcp_…"# then, in Claude Code:> build the Checkout screen from Lindi, in Reactlindi · get_design_context  Checkout · react✓ wrote components/checkout/Checkout.tsx

What your agent
can ask for.

  • Screens and flows.

    The screens in a file, the flows between them, and a picture of any one of them to check its work against.

  • Components and variables.

    The components a screen is made of, and every variable it reads — with its light, dark and brand values.

  • The design context.

    Everything needed to build a frame, in the framework you name: React, HTML and CSS, SwiftUI or Compose.

A screenshot,
or the design itself.

A screenshot

What agents usually get
Pixels to measure
Colours to guess
Components to invent
Spacing to eyeball

Lindi

What your agent gets here
Real sizes and layout
The variables your screens read
The components on the canvas
Auto layout, as layout

Before you
connect it.

Which agents work with Lindi?

Any coding agent that can connect to an MCP server over HTTP — Claude Code, Codex and Cursor among them.

Can the agent change my design?

No. The connection is read-only: your agent can read a file, never edit it.

Can my agent write a push into my code?

Yes. Push from the canvas and your agent reads the change over MCP, then writes it in your codebase — or let Lindi Code write it for you.

How do I end a connection?

Revoke its token in Dev Mode. The connection ends the same moment.

Do I need Lindi Code to use an agent?

No. Lindi Code is Lindi’s own agent; yours works beside it. Either one can take a design from the canvas into your code.

Start coding while you design.

Lindi is free for its early users. Tell us about your team and we answer within 48 hours — yes or no.

  1. 1You connect your project
  2. 2We open your screens with you, on a call
  3. 3Your first push lands in your code