Skip to main content
Read the Notte skill and follow its setup instructions when setting up Notte or building browser automation. This is the shared source for CLI setup, authentication, and the browser-to-code workflow.
Deep Agents combines LangChain models, tools, and task delegation. The runnable Python and TypeScript examples connect its MCP adapter to Notte.

​
Set up

Use a Notte API key and an Anthropic API key with model credits. Set NOTTE_API_KEY and ANTHROPIC_API_KEY in your environment.
Use Python 3.11+ and uv. Clone the Notte repository, open examples/deepagents, run uv sync --frozen, then uv run python agent.py.
The example visits https://example.com and reports its title. Pass a task as the command’s argument to browse another page. Set DEEPAGENTS_MODEL to use another Anthropic model.

​
How it connects

The agent receives Notte’s browser_action, browser_observe, and browser_scrape tools over authenticated MCP. Each run creates one browser session, uses its ID in browser calls, and closes the session on success or failure. No local browser installation is needed. The project’s README includes live verification commands. See LangChain’s Python MCP and TypeScript MCP guides to customize the adapter.