Every credential you own,
behind one endpoint.
Routes on live rate-limit headers, keeps a conversation on the account holding its prompt cache, fails over without dropping a stream, and queues honestly when the whole pool is dry.
-
Cache-aware by default Cache-blind pooling costs 1.94x on the same workload. Affinity first; rebalance only when the owner genuinely cannot serve.
-
Every decision inspectable Each request records why it went where it did — candidates, scores, verdicts. "Load balanced" is not an answer.
-
Accounts you manage here Add, name, test and rotate API keys, Bedrock, Vertex and Claude subscriptions — no config file, credentials encrypted at rest.
Point any Anthropic client at
export ANTHROPIC_BASE_URL= export ANTHROPIC_AUTH_TOKEN=<your key>
Sign in
An admin key unlocks the pool, the usage charts and account management.
Where do I find it?
Any key in your tokenbiryani.yaml marked admin: true. Mint more from the Keys screen once you are in.
keys:
- key: bir_…
name: default
admin: true