Klura

What is Klura?

Klura is an MCP server that wraps the @klura/runtime, exposing browser automation and skill discovery tools to any MCP client (Claude Desktop, Claude Code, Cursor, Windsurf, and others). It lets agents drive websites once, save interaction recipes, and replay them without opening a browser.

How to use Klura?

Install globally via npm install -g @klura/mcp, then add the server definition to your MCP client’s config using the provided JSON snippet with "command": "npx" and "args": ["-y", "@klura/mcp"]. Restart the client; the agent picks up the Klura toolset automatically.

Key features of Klura?

• Browser automation tools: start_session, perform_action, get_screenshot, get_a11y_tree
• Discovery + persistence tools: save_strategy, execute, list_platform_skills, get_strategy
• Network-log inspection and reverse‑engineering escape hatches
• Resource klura://reference served section‑by‑section via URL fragments
• Thin wrapper: each tools/call dispatches to the corresponding klura runtime function

Use cases of Klura?

• Automate repetitive web tasks by recording interactions and replaying them without a browser
• Let agents reverse‑engineer web applications using network logs and JavaScript introspection
• Enable agents to fill forms, scrape data, or navigate sites via the accessibility tree
• Persist and share reusable skill recipes across MCP clients

FAQ from Klura

What dependencies does Klura require?

It requires Node.js and the @klura/runtime package, which is installed automatically as a dependency and auto‑starts a local daemon on first use.

How do I configure my MCP client to use Klura?

Add the server definition to your MCP client’s config using the command npx with arguments ["-y", "@klura/mcp"]. The exact config file location depends on the client; the server definition is the same for all.

Where are skills and data stored?

All learned skills and runtime data are stored under ~/.klura/, with individual skills saved in `~/.klura/skills/