MC-1 OpenAI-compatible Google ADK recipe and streamed tool-loop evidence

[ColomboAI]

Hello Google ADK maintainers,

We are the ColomboAI integration team. We are integrating MC-1 — Auto Intelligence, our Intelligence Control Plane, with agent and developer frameworks.

Models generate intelligence. Agents turn intelligence into action. MC-1 controls the system between them. It is the layer that decides how intelligence is selected, governed, evaluated, and improved. Google ADK therefore gains an adaptive intelligence layer without requiring the framework or its users to choose a permanent model-provider winner.

ADK's existing LiteLlm, LlmAgent, and Runner seams appear sufficient, so we are not proposing a dedicated model class or community extension. Our provider-neutral recipe uses openai/colomboai/mc-1, https://mc1.cairo.sh/v1, and MC1_API_KEY.

The reviewed evidence uses Google ADK 2.7.1 with LiteLLM 1.97.0. The bounded no-network released-host proof, wheel-lock evidence, configuration guide, and known limitations are public. Authenticated MC-1 production validation remains separate and is not presented as complete.

Would a concise generic OpenAI-compatible endpoint recipe be useful in adk-docs, or is another documentation surface preferred? We would welcome guidance before opening any issue, code change, or documentation pull request, and are interested in a deeper documentation and conformance partnership around the existing generic model seam.

Technical overview: https://cairo.sh/MC-1

Regards,
ColomboAI integration team

[surajksharma07]

Nice to see MC-1 working cleanly through the existing LiteLLM OpenAI-compatible seam without needing any ADK-side changes.
since it's a third-party integration showcase rather than an ADK bug or open question. Feel free to keep sharing recipes like this as you build out more of the integration and happy to revisit the docs/conformance partnership question if you want to open a separate discussion for it.