Manifest LLM Gateway

Manifest LLM Gateway connects AI agents and apps to local models, cloud providers and subscriptions through a shared endpoint. It's for builders and teams who want to choose where requests go, handle provider failures and track inference usage.

• Model routing: Connect provider API keys, subscription OAuth tokens and custom OpenAI- or Anthropic-compatible endpoints through one OpenAI-compatible API.
• Request repair and fallbacks: Autofix repairs broken requests before returning an error to your agent, while fallbacks retry calls with another model or provider when a service fails.
• Local model connections: Route calls to Ollama, LM Studio, llama.cpp or another local server, and choose which requests stay on your machine and which go to cloud providers.
• Logs and spending: Inspect full request and response bodies for successful and failed calls, track tokens, and view costs by agent, key and provider.
• Visual route settings: Set models and parameters for each route through the interface without changing application code.

The gateway is free, open source under the MIT license, and available for self-hosting with Docker on your machine or infrastructure. Manifest also offers a hosted cloud version. Local model requests can stay on your machine; calls routed to external providers go to those services and use your API keys or subscription credentials.