> ## Documentation Index
> Fetch the complete documentation index at: https://docs.notte.cc/llms.txt
> Use this file to discover all available pages before exploring further.

# Prime Intellect

> Build Verifiers browser tasksets with Notte sessions.

<Tip>
  Read [the Notte skill](https://notte.cc/skill.md) 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.
</Tip>

[Prime Intellect Verifiers](https://docs.primeintellect.ai/verifiers/v1/overview) provides tasksets, tools, and rewards for agent evaluation and training. The [runnable Python example](https://github.com/nottelabs/notte/tree/main/examples/prime-intellect) gives each rollout an owned Notte browser and exposes a page-title tool through Verifiers' native MCP interface.

## Set up

Open `examples/prime-intellect` in the Notte repository and follow its README. Use Python 3.11+, a [Notte API key](https://console.notte.cc/apikeys), and your configured [inference endpoint](https://docs.primeintellect.ai/inference/overview).

## How it connects

The task opens a three-minute session during setup; the environment closes it in `finally`, including on provider failure. The tool uses Playwright over Notte's CDP connection. Verifiers passes the connection through its authenticated private state channel; the model receives a navigation tool rather than a session endpoint. The reward checks the title and URL observed by the browser.

The example uses the current Verifiers 0.4 taskset API. Its native MCP, reward, real browser, and failure-cleanup tests pass. Paid model rollouts and hosted Prime training remain unverified.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.