/v1/messagesAnthropic Messages API
Claude-compatible messages, streaming, tools, images, and thinking.
Everything included. Nothing metered.
Compatible APIs, account routing, streaming, retries, and every model — with no per-token charge on any request you send.
Token billingNone
Usage includedUnmetered
In-flight workNo concurrency cap
API formatsAnthropic and OpenAI
Use the endpoint your client expects.
The public API supports the request formats used by Claude, Codex, GPT, and compatible tools.
/v1/responsesOpenAI Responses API
Codex and GPT requests through the native Responses shape.
/v1/chat/completionsOpenAI Chat Completions
Compatibility for clients that still use the chat completions format.
Routing that accounts for real capacity.
Policy first
Plan access, model grants, forced-model settings, account health, and current load determine the eligible route.
Bounded retries
Eligible transient failures can move to another account before output begins. Admission failures return immediately.
Direct streaming
Provider output streams from the edge to the client while the control plane keeps credentials and routing policy separate.
Usage data that answers practical questions.
The dashboard focuses on what ran, where time went, and how much of each token type was used.
Live request state
See queued, retrying, completed, and failed work without capturing prompt content.
Token accounting
Track input, output, cache-read, and cache-write tokens separately.
Model analysis
Filter usage by requested model and review daily trends.
Key controls
Create, disable, and filter activity for your own local API keys.
Change the host, not the workflow.
Environment variables cover Claude Code and many agent clients. SDKs can set the same base URL directly.
Environment
export ANTHROPIC_BASE_URL=https://openlimits.app
export ANTHROPIC_AUTH_TOKEN=$OPENLIMITS_KEYTest the full workflow for €10.
24 hours of unmetered access to every model, plus the compatible APIs, routing, streaming, and dashboard analytics.