What is Manifest LLM Gateway?
Manifest LLM Gateway is an open-source LLM gateway for AI agents and apps. Connect your API keys, subscriptions, and local models to one OpenAI-compatible endpoint, and each query goes to the right model. No single-provider lock-in.
- 🔀 Custom Routing: API keys, Subscriptions, Local models, Custom providers
- 💾 Full Body Logs for Success and Error Messages
- 📊 Track every single dollar, setup notifications and limits
- 🚑 Fallback on different models when queries fail, Self-heals your bad requests
Quick start
Cloud version
Go to app.manifest.build and follow the guide.
Self-hosted
Manifest ships as a Docker image. One command:
bash <(curl -sSL https://raw.githubusercontent.com/mnfst/llm-gateway/main/docker/install.sh)
Open http://localhost:2099 and sign up — the first account you create becomes the admin. Full self-hosting guide: docker/DOCKER_README.md.
Deploy with one click
| Platform | Notes |
|---|
| Railway | Best path. Template includes Manifest, PostgreSQL, and S3-compatible storage for request recordings. |
| Render | Blueprint includes Manifest, PostgreSQL, and a persistent recording disk. |
| DigitalOcean | App Platform and PostgreSQL; provide a private Space for recordings. |
| AWS | CloudFormation provisions ECS, RDS, and a private recording bucket. |
| GCP | DeployStack provisions Cloud Run, Cloud SQL, and Cloud Storage. |
Every deployment path now uses durable request-recording storage. Railway, AWS, GCP, and Fly.io provision it natively; Render, Coolify, Easypanel, Docker, and Apple Containers mount persistent storage. DigitalOcean, Heroku, and Koyeb collect external S3-compatible settings during setup. Volume-backed templates are single-instance; use S3-compatible storage before scaling horizontally.
Full deployment guides: Railway, Render, DigitalOcean, AWS, GCP, Fly.io, Coolify, Easypanel, Heroku, Koyeb, and Apple Containers.
The old npm-based self-hosting path is no longer supported. Use the Docker image or one of the deployment guides above.
Providers
Manifest connects to 300+ models through 35 built-in provider connections plus any custom OpenAI/Anthropic-compatible endpoint. Bring your own API key, reuse one of 18 subscription flows, or run models locally. Everything is routed through the same OpenAI-compatible endpoint — send "model": "auto" and Manifest picks the model.
Provider catalogs are discovered dynamically when credentials are connected. The examples below are representative, not exhaustive.
| Provider | API key / local | Subscription | Model catalog |
|---|
| OpenAI | ✅ | ✅ ChatGPT Plus / Pro / Team | GPT-5.6 (Sol / Terra / Luna), GPT-5.5, GPT-5.4, Codex, o-series |
| Anthropic | ✅ | ✅ Claude Max / Pro | Claude Opus 5, Sonnet 5, Fable 5, Haiku 4.5 |
| Google | ✅ | ✅ Sign in with Google | Gemini 3.5 Flash, 3.1 Flash-Lite, Gemini 2.5 |
| Google Vertex AI | ✅ | — | Gemini models via Vertex AI |
| Gemini Free | ✅ Managed key | — | Free Gemini models through Manifest's managed gateway |
| Meta | ✅ | — | Muse Spark 1.2 / 1.1 + Contributor route (Meta Model API) |
| xAI | ✅ | ✅ Grok subscription | Grok 4.5, Grok 4.3, Grok Build, Grok 4.20 |
| AWS Bedrock | ✅ | — | Claude, GPT, Kimi, MiniMax, Nemotron, Nova via Bedrock |
|
Quick links
License
MIT
