Model Context Protocol (MCP) Server

What is the MCP server?

Iterel exposes your living project brain as a Model Context Protocol (MCP) server. This allows external AI coding assistants and chat apps to securely read your project's context, research, strategy, flows, tasks, screens, and virtual file system.

Instead of manually copy-pasting research or design tokens into Cursor or Claude, you can configure them to connect to Iterel. When you ask your IDE to "build a component using my project's design system," it automatically fetches the exact tokens and creative direction from Iterel.

Available capabilities

When connected, external tools have access to the following resources and tools:

• list_projects — List all of your Iterel projects.
• get_project_context — Read the complete living context including user input, research, strategy, and flow phases.
• get_research — Fetch market analysis, personas, and competitors.
• get_strategy — Retrieve your value proposition, positioning, and MVP features.
• get_design_tokens — Access color systems, typography, and design tokens.
• get_screens — List all generated design screens.
• get_tasks — Read tasks from your Kanban board.
• get_flows — Retrieve the user flow graph.
• list_vfs_files and read_vfs_file — Access the generated virtual file system code.

What you can do (Examples)

Once your IDE or chat app is connected to Iterel, you can ask it context-aware questions and tasks without needing to manually copy-paste project details:

• "Use Iterel's get_design_tokens tool to style this React component." Your IDE will fetch the brand colors, typography, and spacing defined by Iterel's Creative Director, and apply them directly to your code.
• "Fetch the research personas from Iterel and write a localized onboarding email." Claude Desktop will pull the exact target audience definitions and write highly tailored copy.
• "Check the Iterel Kanban board tasks and implement the next unassigned frontend task." Cursor will read your project's backlog and automatically start scaffolding out the code required.
• "Review my current codebase against the backend architecture defined in the Iterel flows." Your external AI will retrieve the mapped user flows and architecture plan, identify gaps, and suggest how to align your codebase with the strategy.

Generating an API key

To connect an external tool, you need an Iterel API key:

1. Open Settings (gear icon) in Iterel.
2. Navigate to the MCP Server section.
3. Click New Key and give it a name (e.g. "Cursor").
4. Copy the generated key (it starts with itrl_). You will only see this key once.

Connecting Cursor

Add the Iterel MCP server to your Cursor project by creating or updating the .cursor/mcp.yml file:

---
mcpServers:
  iterel:
    url: https://iterel.com/api/mcp
    headers:
      Authorization: "Bearer <YOUR_ITEREL_KEY>"

Connecting Claude Desktop

Open your Claude Desktop config file (located at ~/Library/Application Support/Claude/claude_desktop_config.json on macOS) and add:

{
  "mcpServers": {
    "iterel": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/inspector",
        "https://iterel.com/api/mcp"
      ],
      "env": {
        "MCP_AUTH_TOKEN": "<YOUR_ITEREL_KEY>"
      }
    }
  }
}

Connecting Zed

Open your Zed settings (~/.config/zed/settings.json) and configure the MCP server similarly to Claude Desktop:

{
  "mcp": {
    "servers": {
      "iterel": {
        "command": "npx",
        "args": [
          "-y",
          "@modelcontextprotocol/inspector",
          "https://iterel.com/api/mcp"
        ],
        "env": {
          "MCP_AUTH_TOKEN": "<YOUR_ITEREL_KEY>"
        }
      }
    }
  }
}

Security and privacy

• Every request is authenticated via your API key.
• The server only exposes projects that your user account has access to.
• It is a read-only bridge. External tools can fetch context but cannot modify your Iterel project data, tasks, or designs.