Claude Code - Pioneer

Pioneer supports a growing list of coding agents. Point your agent at Pioneer’s inference endpoint and switch between models using each agent’s native /model command. Steps to integrate Claude Code with Pioneer:

Download and set up Claude Code.
Open ~/.zshrc (or ~/.bashrc if you use bash) and add these two lines at the bottom to point Claude Code to Pioneer:
```
export ANTHROPIC_API_KEY="<Your Pioneer API key>"
export ANTHROPIC_BASE_URL="https://api.pioneer.ai"
```
Open a new terminal, or run source ~/.zshrc (or source ~/.bashrc) in existing ones to apply the changes.
Start Claude Code
```
claude
```
To integrate Claude Code with our router, access the model picker
```
/model
```
Select pioneer/auto and press Enter to set it as your default for future sessions.
```
pioneer/auto
```

Already logged into Claude Code with a Claude.ai account? If you see the warning Both claude.ai and ANTHROPIC_API_KEY set · auth may not work as expected, Claude Code is still authenticating with your claude.ai login instead of your Pioneer API key.

Run /logout inside Claude Code to sign out of claude.ai. (If the warning persists after relaunching, exit Claude Code and run claude /logout from your terminal instead.
Relaunch with claude. If prompted to log in to claude.ai, decline — when asked to use the ANTHROPIC_API_KEY from your environment, accept.
Verify with /status:
- Auth token should show none
- API key should show ANTHROPIC_API_KEY
- Anthropic base URL should show https://api.pioneer.ai.

Still seeing the warning? On macOS, open Keychain Access, search for “Claude Code”, delete the stored credentials entry, and relaunch.

Prompt caching works automatically. Claude Code talks to Pioneer over the Anthropic Messages API and sends cache_control markers itself, so prompt caching is applied on Claude models with no extra setup. See Prompt Caching.