Back to guides

Oh My Pi Integration

Connect Oh My Pi to LLM Gateway or DevPass, discover models, and run a coding task from your terminal.

Oh My Pi (omp) is a terminal coding agent with file editing, shell, language-server, and debugger tools. Connect it to LLM Gateway with a pay-as-you-go or DevPass API key.

Video walkthrough

Install Oh My Pi

On macOS or Linux:

1curl -fsSL https://omp.sh/install | sh2omp --version

See the installation guide for other platforms and package managers.

Connect LLM Gateway or DevPass

Create an API key in your LLM Gateway dashboard or DevPass dashboard. Export it in the terminal where you will run Oh My Pi:

1export LLMGATEWAY_API_KEY="your_api_key"

Create or update ~/.omp/agent/models.yml. Merge this entry into any existing providers mapping:

1providers:2  llmgateway:3    baseUrl: https://api.llmgateway.io/v14    api: openai-completions5    apiKey: LLMGATEWAY_API_KEY6    discovery:7      type: openai-models-list

apiKey names the environment variable. Discovery loads the gateway's /v1/models catalogue, so you do not need to maintain a model list in the configuration. The same endpoint works with either kind of key; the key determines your workspace and billing.

Select a model

Start Oh My Pi in your project:

1cd your-project2omp

On first launch, finish the setup wizard; the configured provider appears in the model step.

Run /model, choose llmgateway, and select a text model with tool support from the live catalogue. With DevPass, choose a model supported by your plan.

To select a model when launching, replace MODEL_ID with its canonical catalogue ID:

1omp --model llmgateway/MODEL_ID

Here, llmgateway/ selects the custom provider inside Oh My Pi. It is not an upstream provider pin. Keep the remaining model ID canonical for DevPass routing.

Verify the connection

Ask Oh My Pi to fix a small failing test in your project:

1Read the failing test, make the smallest fix, run the tests again,2and summarize the result. Do not edit the tests.

Review the diff and test output, then check usage in the dashboard for the workspace that issued your key.

For a one-shot task:

1omp --model llmgateway/MODEL_ID -p "Explain this project without changing files"

Troubleshooting

  • Provider missing: Check ~/.omp/agent/models.yml, including YAML indentation, then restart omp.
  • Authentication failed: Export LLMGATEWAY_API_KEY in the same shell before launching the agent.
  • Model rejected: Select a tool-capable model available to your account. DevPass does not support upstream provider-pinned model IDs.
  • Discovery failed: Check the base URL and use openai-models-list for the gateway's OpenAI-compatible catalogue.

See the Oh My Pi model configuration reference for additional settings.