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

# Cursor

> Configure Cursor's chat panel to use Pioneer's OpenAI-compatible API, add custom models, and switch between them from the model picker.

Steps to integrate Cursor with Pioneer:

## Setup

<Warning>
  This integration only covers Cursor's **chat / plan panel** (`Cmd/Ctrl+L`). Composer, inline edit (`Cmd/Ctrl+K`), tab completion, and apply/edit are locked to Cursor's own backend and cannot be routed to Pioneer. If you need a full coding agent backed by Pioneer, use [Claude Code](/claude-code) or [Codex CLI](/codex) instead.
</Warning>

<Steps>
  <Step title="Open the Cursor desktop app">
    Make sure you're using the **Cursor desktop app** (the code editor, with a file tree and sidebar) — not the Cursor CLI agent (`agent` command) and not the [cursor.com](https://cursor.com) web account dashboard. Neither of those has the Models settings described below. If you only have the Cursor CLI agent installed, download the desktop app from [cursor.com](https://cursor.com).
  </Step>

  <Step title="Open Cursor Settings">
    Press `Cmd+,` (macOS) or `Ctrl+,` (Windows/Linux), or click the gear icon in the top-right toolbar. You can also use `Cmd/Ctrl+Shift+J`.

    Make sure you land on **Cursor Settings**, not **VS Code Settings** — only Cursor Settings has a Models tab.
  </Step>

  <Step title="Go to Models">
    In the Cursor Settings sidebar, click **Models**. Scroll down to the **API Keys** section and expand it.
  </Step>

  <Step title="Set your Pioneer key and base URL">
    Under the OpenAI provider section:

    * Enable the **OpenAI API Key** toggle and paste your Pioneer API key
    * Enable **Override OpenAI Base URL** and enter:

    ```text theme={null}
        https://api.pioneer.ai/v1
    ```

    Save / verify. Cursor will send a test request.
  </Step>

  <Step title="Add your Pioneer model">
    Back in the Models list, click **+ Add Custom Model** and enter this exact model name:

    ```text theme={null}
        pioneer/auto-claude-opus-4
    ```

    Toggle it on, then click **Verify** / save.

    <Warning>
      Type the model name exactly as `pioneer/auto-claude-opus-4`, character for character. This is the Pioneer autorouter alias — it still routes across Claude, GPT, and DeepSeek — but the embedded `claude-opus-4` slug is required: Cursor only emits Anthropic `cache_control` prompt-caching markers when the model name contains a recognized Claude model slug (a real family plus numeric major version), and the autorouter depends on those markers to cache prompts. Renaming it, dropping the version, or using a plain `pioneer/auto` name silently disables prompt caching.
    </Warning>
  </Step>
</Steps>

## Using the model

Open the chat/plan panel with `Cmd+L` (or `Ctrl+L`), then click the model picker dropdown at the bottom of the panel (where it shows the current model, e.g. "Auto" or "Sonnet 4.5") and select your Pioneer model from the list.

## Troubleshooting

<AccordionGroup>
  <Accordion title="I don't see a Models tab in Settings" icon="settings">
    You're likely in the wrong app. The Cursor CLI agent's settings (`~/.cursor/cli-config.json`) and the cursor.com web account dashboard (Plan & Usage, Active Sessions, etc.) both lack a Models tab. Open the Cursor **desktop app** itself and use `Cmd+,` from inside it.
  </Accordion>

  <Accordion title="Invalid Model or network errors after setting Override Base URL" icon="wifi-off">
    Double-check that the base URL includes `/v1` and that your Pioneer API key is correct and active. If errors persist, try disabling other enabled models in the Models list to rule out conflicts, then re-enable only your Pioneer custom model.
  </Accordion>
</AccordionGroup>