Skip to main content
The Aixy extension for Pi discovers your project’s models and sends conversations, streaming output, and function-tool calls through Aixy’s Chat Completions API. Use Pi 0.87.0 or newer from @earendil-works/pi-coding-agent and Node.js 22.19 or newer.

​
Install and sign in

Restart Pi or run /reload. Enter /login aixy, supply your Aixy project key, then use /model to select an Aixy model. Choose a model with function-tool support for coding tasks. Alternatively, use an environment key:
Replace the example ID with one listed for your project. Routing aliases work unchanged, for example pi --provider aixy --model aixy/coding after creating that route.

​
Configure a custom gateway

This setting changes both discovery and inference. Keep the key aligned with that gateway. Pi’s saved login takes precedence over AIXY_API_KEY; use /login aixy to replace it or /logout aixy before switching to an environment credential.

​
Verify and maintain

Send a short prompt and inspect the resulting project request in Activity. Model discovery supplies IDs; configure model limits and capabilities in ~/.pi/agent/models.json using verified values. Pi’s initial local cost defaults are not evidence of free inference. See the extension’s model configuration instructions. Update the extension with:
Restart or reload Pi afterward. Keep its credentials and model configuration outside the repository. For shared tool accounts, configure MCP access separately in a client setup that supports remote MCP.