Agents & Editors
Pi.Dev
Connecting the pi coding agent to an Adaptive Router with the OpenAI Chat Completions format.
pi is a coding agent for the terminal that works with many model providers. It connects to Multi-Router as a custom provider in its models file, sending OpenAI Chat Completions requests to an Adaptive Router, so one setup reaches every model the router serves.
Creating the router
pi connects through an Adaptive Router, which accepts OpenAI Chat Completions requests and serves each one with the model it names, using the organization's provider key for that model's provider. pi holds only the Multi-Router API key. See Adaptive Router. To set it up:
- In Model Groups, create or choose a model group with the models pi should use. See Model Group.
- In Routers, choose New router, then Adaptive Router. Name it, choose the model group, and switch on the models to serve.
- Choose the organization's provider key for each provider in the group, and create the router. Its id, such as
rt_XXXXXXXX, is on the router's page. - In API Keys, create a key with the
inferencescope. It can be limited to this router. See API Key.
Every request carries the router id in the X-MR-Router header and the API key as a Bearer token, both set in pi's settings below.
Settings
pi reads custom providers from ~/.pi/agent/models.json. Add a provider that points at Multi-Router, uses the openai-completions API and sends the router header. apiKey and header values accept $NAME to read an environment variable, so the key stays out of the file:
multi-router is a local name for the provider; any name works. pi reloads the file when /model is opened.
Models
An Adaptive Router names models by their Multi-Router catalog id, provider/model, such as google/gemini-2.5-flash, openai/gpt-5 or google/gemini-2.5-flash. The model must be switched on in the router's model group; a request for any other model is refused with a message listing the models the router serves. The settings on this page list each model id explicitly.
Choose a model with /model inside pi, or at launch with --provider multi-router --model google/gemini-2.5-flash..
Checking the setup
The request then appears in Traffic with the Adaptive Router in the Router column and the model, tokens and cost in the request detail. A reply with the requested model or router was not found means the router id is wrong, or the API key is not allowed to use that router. See Traffic and Troubleshooting.