mnfst/manifest

What it solves

Manifest is an open-source LLM gateway designed to prevent AI agents and applications from breaking due to provider lock-in or service failures. It provides a single, OpenAI-compatible endpoint that abstracts away the complexity of managing multiple API keys, subscriptions, and local models.

How it works

Users connect their various AI providers (including API keys, subscriptions like ChatGPT Plus or Claude Pro, and local runners like Ollama or LM Studio) to Manifest. The gateway then routes queries to the appropriate model. By sending a request with "model": "auto", Manifest automatically selects the best model for the query.

Who it’s for

It is built for developers of AI agents and applications who want to avoid dependency on a single LLM provider and need a reliable way to manage costs, logs, and model fallbacks.

Highlights

• Broad Provider Support: Connects to over 300 models via 32 built-in provider connections, including major labs and local inference engines.
• Reliability Features: Includes automatic fallbacks to different models when queries fail and self-healing for bad requests.
• Unified Interface: Offers an OpenAI-compatible endpoint for seamless integration.
• Observability: Provides full body logs for success and error messages, and tools to track spending, set notifications, and establish limits.
• Flexible Deployment: Supports self-hosting via Docker or one-click deployments on platforms like Railway, AWS, GCP, and Render.