The Open Source LLM Gateway
Route 280+ models across 35+ providers through one OpenAI-compatible API — and run the entire platform on your own infrastructure. Open source, self-hostable, no lock-in.
Why an open source AI gateway matters
The infrastructure routing every model call is too important to be a black box.
No vendor lock-in
The whole platform is AGPLv3. Fork it, audit it, and run it forever — no proprietary control plane you can be cut off from.
Self-host the full stack
Gateway, dashboard, and worker ship in a single Docker image. Keep every request and key inside your own infrastructure.
Data residency by default
For regulated and privacy-sensitive teams, requests never have to leave your network or pass through a third party.
Inspect and extend
Read the code, open a PR, or bend it to your stack. An LLM API gateway you can actually change beats one you can only call.
Self-host in one command
The gateway, dashboard, and worker in a single image.
docker run -d \
--name llmgateway \
-p 3002:3002 -p 4001:4001 -p 4002:4002 \
-e AUTH_SECRET="your-secret" \
-e GATEWAY_API_KEY_HASH_SECRET="your-hash-secret" \
ghcr.io/theopenco/llmgateway-unified:latestPrefer not to run infrastructure? The managed LLM API gateway is pay-as-you-go, or free with your own provider keys.
Open vs closed gateways
Most gateways open-source a router at best. LLM Gateway open-sources the whole platform.
| Gateway | Open-source scope | Self-host |
|---|---|---|
| LLM Gateway | Full platform (AGPLv3) | |
| OpenRouter | Closed source | No |
| Vercel AI Gateway | Closed source | No |
| Cloudflare AI Gateway | Closed source | No |
| Portkey | Gateway + parts (MIT) | Partial |
| LiteLLM | Library/proxy (MIT) |
FAQ
Open source LLM gateway FAQ
Licensing, self-hosting, and how the open-source build relates to the managed service.
Can't find an answer? Contact us