What it does

Manifest LLM Gateway is an open-source gateway for AI agents and apps. You connect API keys, subscriptions and local models, and your application talks to a single OpenAI-compatible endpoint. Sending "model": "auto" lets Manifest choose the model, which avoids tying an app to a single provider. The README lists 35 built-in provider connections, 18 subscription sign-in flows and access to more than 300 models, plus any custom endpoint that follows the OpenAI or Anthropic request format. Local servers such as Ollama, LM Studio and llama.cpp are supported. The gateway keeps full request and response bodies for successes and errors, tracks spend so you can set notifications and limits, and falls back to other models when a query fails. It also describes repairing malformed requests on the fly. You can use the hosted cloud version at app.manifest.build or self-host with a Docker image installed by one command, where the first account created becomes the admin. One-click deployment guides cover Railway, Render, DigitalOcean, AWS, GCP, Fly.io and others, and request recordings need durable storage. An n8n community node is also published. The repository is MIT licensed.

Capabilities

One OpenAI-compatible endpoint across keys, subscriptions and local models
Automatic model selection with "model": "auto"
Full request and response body logs
Spend tracking with notifications and limits
Fallback to other models when a request fails
Docker self-hosting plus a hosted cloud version

When to use it

You want to route agent traffic across several providers without changing application code
You want to see every request and what it costs
You want to mix cloud models and local models behind one endpoint
You need to host the gateway on your own infrastructure