Compatible (BYOK)

Use Pi with GlideflowAI.

A minimal terminal coding harness with documented custom OpenAI-completions providers. Connect it through GlideflowAI's OpenAI-compatible API with a key you control.

Get a key

What this connection is for

Pi is a minimal terminal coding harness that can be extended with skills, TypeScript extensions, and custom model providers. It is aimed at developers who prefer a small core and are comfortable declaring exactly how a provider and model should behave. The local configuration can describe providers independently from the agent's extensions and project workflow, which keeps model routing explicit.

Choose Pi when you want to tune a provider definition or build on its extension model without adopting a larger agent platform. It is not the simplest option for someone who only needs a graphical provider form. Because Pi lets a custom provider define API behavior and model metadata, validate a short task before applying the same provider configuration to more complex skills or tools.

Pi's official custom-provider documentation supports OpenAI Completions-compatible providers with baseUrl, apiKey, and model definitions. Register GlideflowAI in ~/.pi/agent/models.json, reference the key through Pi's environment expansion, and select an exact catalog model. Keep provider compatibility flags conservative until a real Pi task confirms the selected model's request behavior. Pi is an independent project; this is a user-managed connection, not an endorsement or partnership.

Why pair it with GlideflowAI

Pi can register GlideflowAI as a named openai-completions provider, making its base URL and model choices visible in the agent's own configuration.

The authenticated model list is helpful when you are defining the provider's available models instead of selecting from a fixed built-in catalog.

Setup

Use a new API key for the tool, select a model from the catalog, and begin with a small task. Do not put a key in a repository, screenshot, or shared configuration file.

1
Set a private key
Use Pi's supported environment-variable expansion rather than placing a live key directly in models.json.
```
export GLIDEFLOW_API_KEY="sk-your-key"
```

Register an OpenAI-completions provider

Add the provider to ~/.pi/agent/models.json and choose the exact model ID through Pi's model selection flow.

{
  "providers": {
    "glideflowai": {
      "baseUrl": "https://api.glideflowai.com/v1",
      "api": "openai-completions",
      "apiKey": "$GLIDEFLOW_API_KEY",
      "models": [{ "id": "glm-5.2" }]
    }
  }
}

Need the endpoint basics? Read authentication, browse the model catalog, or compare pricing.

Frequently asked questions

Which Pi API mode should GlideflowAI use?

Set api to openai-completions in the custom provider because GlideflowAI exposes the OpenAI-compatible Chat Completions shape at /v1.

Where does Pi load this provider definition?

Use Pi's documented ~/.pi/agent/models.json custom-models configuration, then select the provider through Pi's own model flow.

Can Pi read an environment variable in apiKey?

Yes. Pi's custom-provider configuration supports $ENV_VAR expansion, so use $GLIDEFLOW_API_KEY instead of a literal key.

Is this an official integration?

No. The connection is configured by you using a compatible API and does not imply sponsorship, endorsement, or a partnership.

Ready to connect Pi?

Create a key, use the setup above, and verify the connection on a small task before you scale it up.