A managed LLM gateway for your backend services
One endpoint for every model. Your keys. No outages.
router0 sits between your services and your LLM providers. Send a model name to a single OpenAI-compatible API. router0 picks the provider, fails over when something breaks, and tells you plainly why when nothing can serve the request.
Works with the providers you already use
- Anthropic
- OpenRouter
- Groq
- Google Gemini
- Mesh API
Calling several LLM providers directly gets messy fast
A key per provider, in every service
Each backend carries its own set of provider keys, and rotating one means touching all of them.
A provider hiccup becomes your outage
Rate limits, empty credits and flaky endpoints each need their own retry logic, copied from app to app.
Errors that say nothing
When a call fails you have to guess whether the model is missing, the key is out of credit, or you are throttled.
Everything between your code and the model
Chat completions, streaming and tool calling, behind one API that behaves the same whichever provider ends up answering.
Call a model by name
Send a canonical model name or one of your own aliases. router0 works out which provider can serve it, so callers never hard-code a vendor.
Routing you control
Set a fixed provider order per organization and a key order within each provider. Override it for a single request with the X-Router0-Providers header.
Failover on any error
If a provider or key fails, router0 moves to the next one. Streaming fails over too, right up until the first byte is sent to your client.
Errors you can act on
404 model_not_found, 402 insufficient_credits and 429 rate_limited are separate, with a Retry-After header when you are throttled.
Bring your own keys
Your provider keys stay yours and are stored encrypted. Spend lands on your own provider accounts, not ours.
Teams and service tokens
Organizations hold many named, revocable service tokens. Give each backend its own and cut one off without touching the rest.
See every request
A request log shows the model, status, latency and the full failover trail: which providers were tried, in what order, and what each said.
Live model catalog
Models are discovered from your providers and kept fresh. Browse them in the dashboard, search by kind, and define tenant-level aliases.
Up and running in three steps
- 1
Add your provider keys
Paste keys for the providers you already use. router0 checks them and lists the models each one offers.
- 2
Set your routing order
Choose which provider is tried first and how your keys are ordered within it.
- 3
Point your app at router0
Create a service token and swap the base URL. Your existing OpenAI-compatible client keeps working.
Change the base URL. That is the migration.
router0 speaks the OpenAI chat completions API, including server-sent-event streaming and tool calling. Agents, SDKs and frameworks that let you set a base URL work without code changes.
Model names are canonical (or your own aliases), so the same request works regardless of which provider serves it.
curl https://api.router0.smalltech.in/v1/chat/completions \
-H "Authorization: Bearer $ROUTER0_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"model": "llama-3.3-70b",
"messages": [{"role": "user", "content": "Hello!"}]
}'from openai import OpenAI
client = OpenAI(
base_url="https://api.router0.smalltech.in/v1",
api_key="r0-sk-...", # your router0 service token
)
resp = client.chat.completions.create(
model="llama-3.3-70b",
messages=[{"role": "user", "content": "Hello!"}],
)Stop wiring up providers. Start shipping.
router0 is invite-only for now. If you have an invite, create your organization and your first token in a couple of minutes.