Line9's renderer, inside your assistant.

Line9 renders the Mermaid you already write with its own layout engine — same .mmd source, cleaner placement, orthogonal edges, readable type. Until now that lived in the web viewer and the CLI. The hosted MCP server puts it in the conversation: ChatGPT, Claude, Grok.

Before, and after

An assistant writes Mermaid and the host draws it with stock layout — overlapping nodes, tangled edges, a diagram you would not put in a doc. Add Line9 and the same source is laid out by Line9's engine and shown as a zoomable widget in the thread, one click from line9.ai/diagram.

Hosts

Connect from claude.ai and Claude Desktop, ChatGPT (developer mode), and Grok. Working in a terminal or code editor, with Claude Code, Codex, Cursor or similar? Use the Line9 CLI instead: line9 render opens the diagram in a browser tab and updates it as your agent edits, and line9 bootstrap tells the agent Line9 is there.

Add Line9

Every host takes the URL https://mcp.line9.ai/mcp — steps for each are on the MCP docs.

Apps that ask for a config file take this:

{
  "mcpServers": {
    "line9": {
      "url": "https://mcp.line9.ai/mcp"
    }
  }
}

Try it

In Claude, say "using Line9" — otherwise it may draw with its own renderer.

  • Using Line9, draw a flowchart of how a pull request gets merged
  • Using Line9, turn this process into a diagram: …
  • Show me my Line9 diagrams

Data handling

How Line9's MCP handles your data is written out once, on the docs — not restated here.

More

Full connect instructions, tools, sign-in and troubleshooting: /docs/mcp. Commercial use and Pro: pricing.