@lando-labs/cami-mcp

The CAMI MCP server. Manage AI agents from any IDE that speaks the Model Context Protocol.

CAMI is an agent lifecycle management platform — create, deploy, observe, and maintain AI agents across tools and projects. This package runs CAMI as an MCP server inside your editor.

Install

The recommended setup is via Claude Code's MCP CLI:

claude mcp add cami -- npx -y @lando-labs/cami-mcp

You'll need a CAMI API key. Generate one at your CAMI app's Settings page.

After install, set the key:

claude mcp env cami CAMI_API_KEY=cam_usr_...

Manual config

If you'd rather edit JSON, the equivalent mcpServers block:

{
  "mcpServers": {
    "cami": {
      "command": "npx",
      "args": ["-y", "@lando-labs/cami-mcp"],
      "env": {
        "CAMI_API_KEY": "cam_usr_..."
      }
    }
  }
}

Drop this in:

• Claude Code — ~/.claude/settings.json (or your project's .claude/settings.json)
• Cursor — ~/.cursor/mcp.json
• Other MCP-capable clients — wherever they expect MCP server config

Restart your editor after editing.

Verify

In your editor, ask:

Hey CAMI, are you there?

If everything's wired, CAMI's home view opens.

Tools

• cami_home — overview of your workspace
• cami_list_agents / cami_get_agent — browse agents
• cami_create_agent / cami_update_agent — create or revise agents via the architect
• cami_validate_agent — validate an agent definition
• cami_deploy_agent — deploy an agent to one or more projects
• cami_scout_project — propose a starter agent roster for a project
• cami_open — open any CAMI view in the iframe

Docs

Full documentation: https://github.com/lando-labs/cami-app

License

MIT