# Cursor

> Override Cursor's OpenAI base URL to route chat through Tchavi — and what that path does not cover.

- Canonical: https://tchavi.com/en/docs/coding-agents/cursor

---


Part of the [coding agents](/docs/coding-agents) guides.

Cursor accepts an OpenAI-compatible endpoint, with real limits that are worth reading before you set it up.

## Configure it

1. Open **Cursor Settings → Models**.
2. Under the OpenAI section, paste your Tchavi key (`sk-tch-…`).
3. Turn on **Override OpenAI Base URL** and enter `https://tchavi.com/api/v1`.
4. Add the model ids you want by name — `claude-sonnet-4-6`, `gpt-5.4`, `deepseek-v4-pro`. Cursor does not read the catalogue, so each one is typed in by hand.
5. Click **Verify**. Cursor sends a test request; if it fails, the key or the URL is wrong.

## What this path does and does not cover

<Callout type="warning" title="Read this before relying on it">

- **Chat only.** Tab completion keeps using Cursor's built-in models whatever you configure here, and some agent features require Cursor's own backend.
- **Requests still pass through Cursor's servers,** which build the final prompt before it reaches Tchavi. Cursor's zero-data-retention policy does not apply to requests made with your own key.
- **Teams and Enterprise plans are still charged** Cursor's per-token rate on third-party requests, on top of the Tchavi credits.

</Callout>

Cursor may also reject a base URL that is not publicly reachable, so a local tunnel in front of Tchavi will not work — point it at `https://tchavi.com/api/v1` directly.

If you want the whole agent loop running on Tchavi rather than just the chat pane, a terminal agent is the better fit: [Claude Code](/docs/coding-agents/claude-code), [OpenCode](/docs/coding-agents/opencode), [Deep Code](/docs/coding-agents/deep-code) or [Codex](/docs/coding-agents/codex).

---

Back to [coding agents](/docs/coding-agents)