manifest

Why we included this project

Once your agents talk to several LLM providers, managing a different API key and integration for each one gets old fast. Manifest puts a single OpenAI-compatible endpoint in front of all those calls, so your code stays simple while the gateway routes each request to the right model or subscription. It handles bring-your-own keys, flat-rate plans, and self-hosted models side by side, and can pick the cheapest or most reliable option per query. It also logs full request and response bodies, tracks spend to the dollar with alert thresholds, and switches to a fallback model when a call fails, so pipelines keep running without someone watching. If you prefer to run it yourself, there's a Docker image and one-click deploys for the major clouds; a hosted version is there for teams that want less to maintain.