Unlock

Paste your ADMIN_TOKEN. Stored only in this browser.

Key pool

Every key is an upstream provider account. The router spreads calls across them and cools any key that reports a limit. Secrets are never sent to the browser in full.

provider / labelstatustodayminuteok / errsecret

Add a provider key

Tip: add the same provider several times with different accounts to multiply your free budget.

Playground

—

Egress IP

Production runs with no VPC, so the function exits via AWS's shared NAT pool — a datacenter IP that changes between invocations. Add a VPC + NAT when you want a single static IP.

Recent calls

timeprovidermodeltokensms

Users & API keys

Users authenticate to /v1 with an sk-rt-… key. A key is shown once at issue time; only its hash is stored.

—
userrolekeysrequestslimits

Async jobs

For replies slower than CloudFront's ~60s sync cap. Create a job, it returns 202, a worker finishes it, you poll. Use POST /v1/jobs from your own code.

jobstatusmodelcreatedresult

Use it

Create a user in the Users tab, copy their one-time sk-rt-… key, then point any OpenAI-compatible client at with it.

curl -s /v1/chat/completions \
  -H "Authorization: Bearer $ROUTER_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"auto","messages":[{"role":"user","content":"hi"}]}'

Python: OpenAI(base_url=..., api_key=ROUTER_KEY). Every reply carries x-router-* headers naming the key, provider, attempt count and latency.

Admin endpoints live under /admin (keys, providers, stats, config, egress, logs).