Pi

Pi is a lightweight command-line coding agent. Connect it by adding a custom provider in models.json.

• Protocol: OpenAI-compatible (Chat Completions)
• Base URL: https://tokenfunc.com/v1
• Models: any, for example claude-sonnet-4-5

1. Install Pi

   macOS / Linux

   curl -fsSL https://pi.dev/install.sh | sh

   Windows

   powershell -c "irm https://pi.dev/install.ps1 | iex"

   npm

   # Requires Node.js 22.19 or later
   npm install -g --ignore-scripts @earendil-works/pi-coding-agent

   Run pi --version to confirm. On Windows, install Git for Windows first.

2. Add a custom provider

   Put the following into Pi's model config file, creating it if needed:

   ~/.pi/agent/models.json

   {
     "$schema": "https://pi.dev/schemas/models.schema.json",
     "providers": {
       "tokenfunc": {
         "baseUrl": "https://tokenfunc.com/v1",
         "api": "openai-completions",
         "apiKey": "sk-xxxxxxxxxxxxxxxx",
         "models": [
           { "id": "claude-sonnet-4-5", "contextWindow": 200000, "maxTokens": 32000 }
         ]
       }
     }
   }

   • Set contextWindow and maxTokens to the model's real limits.
   • Add more entries to models to use several models.
   • apiKey can also be "$TOKENFUNC_API_KEY" to read it from an environment variable and keep it out of the file.

3. Choose a model and verify

   # List the models under this site's provider
   pi --list-models tokenfunc
   
   # Send a test message
   pi --model tokenfunc/claude-sonnet-4-5 -p "reply ok"

   A reply means you are connected. Day to day, type /model in Pi to pick a model and press Ctrl+S to make it the default.

Troubleshooting