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

# Use Cursor with abliteration.ai

> Connect Cursor to abliteration.ai with Cursor's OpenAI-compatible API key and base URL settings.

**To use [Cursor](https://cursor.com) with abliteration.ai**, add an Abliteration model to Cursor, enter your `ak_...` key as the OpenAI API key, and override the OpenAI base URL with `https://api.abliteration.ai/v1`.

<Warning>
  Cursor requires Pro, Pro+, or Ultra to use your own API key on an individual account. The Hobby plan only sends Agent requests with **Auto** and blocks a custom named model before sending a request to abliteration.ai. See [Cursor pricing](https://cursor.com/help/account-and-billing/pricing). On a team account, the administrator must also permit personal API keys.
</Warning>

## Get an API key

Create an `ak_...` key in the [console](https://abliteration.ai/console).

## Configure Cursor

1. Open **Cursor Settings**.
2. Select **Models**, then **API Keys**.
3. Under **OpenAI API Key**, enter your `ak_...` key and enable **Use OpenAI API Key**.
4. Enable **Override OpenAI Base URL** and enter `https://api.abliteration.ai/v1`.
5. Return to **Models**. In **Add or search model**, add `abliterated-model-large-v2` and leave it enabled.
6. Start a new Agent chat and select `abliterated-model-large-v2` from the model picker. Do not select **Auto**.

Large V2 is the recommended text-only coding model and has a 1M-token context window. If a task includes images, add and select `abliterated-model` instead. See [Models](/models) for the complete capability comparison.

<Note>
  Cursor applies **Override OpenAI Base URL** to its OpenAI API-key integration. Disable the override before using an OpenAI key directly with OpenAI models.
</Note>

## Verify the connection

Send a short prompt in a new Agent chat. To verify a complete tool loop, use:

```text theme={"system"}
Use the terminal to run pwd, then reply with only DONE.
```

Review and approve the command if Cursor asks. A successful run executes `pwd` and renders `DONE`. You can confirm the request in the [console](https://abliteration.ai/console).

## What this setup covers

* Cursor sends Agent and chat requests through abliteration.ai's OpenAI-compatible Chat Completions API.
* Requests consume your abliteration.ai balance. Your Cursor subscription is separate.
* Cursor's custom API key setting applies to chat models. Features such as Cursor Tab continue to use Cursor's built-in models.
* Published policies scoped to the Abliteration key's project apply to requests from Cursor. See [Policies](/policy-gateway/overview).

Cursor states that [custom API keys pass through its backend for prompt construction](https://cursor.com/help/models-and-usage/api-keys) and that Cursor's zero-data-retention policy does not apply to those requests. Review Cursor's current data-handling terms before sending sensitive code.

Cursor documents its OpenAI API-key support for standard chat models. Abliteration's Chat Completions integration has completed native Cursor Agent tool calls, but Cursor can change custom-model behavior between releases. If Cursor rejects the model before sending a request, check Cursor's current [API-key support](https://cursor.com/help/models-and-usage/api-keys) and plan requirements.

## Troubleshooting

### Cursor says only Auto is available

The current account does not support custom named models. Upgrade to an eligible Cursor plan or switch to an eligible team account. This rejection happens inside Cursor, so no request reaches abliteration.ai.

### The API key controls are unavailable

A team administrator can disable personal API keys. Ask the administrator to allow them, or use an account where the setting is available.

### Authentication fails

Confirm that the OpenAI API key field contains an active `ak_...` key and that the base URL is exactly `https://api.abliteration.ai/v1`.

### Cursor rejects a local base URL

Cursor may reject loopback or private-network API URLs. Use the public HTTPS endpoint `https://api.abliteration.ai/v1`.

### Image input returns `400`

Large V2 is text-only. Add and select `abliterated-model` for prompts that include images.


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