@klura/mcp

MCP (Model Context Protocol) server for @klura/runtime. Exposes the klura runtime's browser-automation + skill-discovery toolset to any MCP client — Claude Desktop, Claude Code, Cursor, Windsurf, and others.

@klura/runtime is the runtime that turns websites into reusable skills: drive the site once, save the recipe, skip the browser on every subsequent run. @klura/mcp is the thin wrapper that speaks MCP on top of it. See the runtime README for what the skills actually do and how they get saved.

Install

npm install -g @klura/mcp

The @klura/runtime package is declared as a dependency and installed alongside. The runtime auto-starts a local daemon on first use and stores everything it learns under ~/.klura/.

Wire it up

Add the server to your MCP client's config. The exact file path depends on the client — the server definition is the same.

{
  "mcpServers": {
    "klura": {
      "command": "npx",
      "args": ["-y", "@klura/mcp"]
    }
  }
}

Restart the client. The agent picks up the klura toolset automatically.

What it exposes

Two surfaces land in the agent's context:

• Tools — browser automation (start_session, perform_action, get_screenshot, get_a11y_tree), discovery + persistence (save_strategy, execute, list_platform_skills, get_strategy), network-log inspection (get_network_log, find_in_page), and the reverse-engineering escape hatches (inspect_ws_frame, try_generator, js_eval, get_js_source, search_js_source, read_js_function, set_breakpoint, wait_for_pause, and more). The runtime owns the canonical list; this server mirrors it one-for-one.
• Resource klura://reference — the detailed reference doc, served section-by-section via URL fragments (klura://reference#reverse-engineer-playbook, klura://reference#recorded-path-schema, etc.) so each response fits inside the MCP output budget. Fetch klura://reference with no fragment for a table of contents.

The always-loaded orientation is SKILL.md, passed as the server's instructions capability. Agents read SKILL.md on every conversation and pull detail on demand via the klura://reference fragments.

How it works

Thin wrapper — each MCP tools/call dispatches to the corresponding klura runtime function. The runtime handles daemon lifecycle, browser sessions, strategy persistence under ~/.klura/skills/, and execution.

MCP client (Claude Desktop, Cursor, …)
    │  stdio transport, JSON-RPC
    ▼
@klura/mcp (this package)
    │  Node require('@klura/runtime')
    ▼
klura runtime → local daemon → Playwright

License

BUSL-1.1. See LICENSE. Same terms as the underlying @klura/runtime; see the runtime README for the commercial-use terms.