Your agent doesn't need ten API keys.
It needs one line.
Point any OpenAI- or Anthropic-compatible agent at Ainfera, set model="ainfera-inference", and every call routes to the model most likely to finish the task — neutral across providers, settled per call, signed on chain. No model-pinning, no ten-vendor key sprawl.
# pip install openai · key from `ainfera install`
client = OpenAI(base_url="https://api.ainfera.ai/v1", api_key=AINFERA_KEY)
res = client.chat.completions.create(model="ainfera-inference", messages=msgs)One key. No pinning. Settled per call.
A single ainfera_ key
One ainfera_ key for the whole fleet — scoped per agent, with caps and a pause toggle on the agent, not a drawer of provider keys to rotate.
You don't choose the model
Send model=ainfera-inference and we route within your caps to whatever finishes the task. Pin a specific model any time to opt out — we still settle and audit it.
Per-call x402 settlement
Each call settles on its own over x402 — sub-$0.001 granularity, no monthly invoice to reconcile. Pricing: /pricing.
Routed to what finishes it
The route is decided per call on the likely outcome, then signed on chain. The mechanism stays ours; the result is yours to verify — how routing works →
Built to be driven by software, headless.
MCP skill
Reachable as an MCP server at mcp.ainfera.ai — your agent discovers and calls Ainfera as a tool, no glue code.
Works without a UI
Sign up, mint a key, call, and verify — all over the API. No dashboard required to operate; the dashboard is for humans who want to watch.
Every surface has a twin
The catalog, the index and the audit trail are all JSON or plain text — /v1/models, /llms.txt, /v1/audit/public.
Drop into the framework you already run.
OpenClaw & Hermes Agent
First-class adapters: route every step through ainfera-inference with one config line. Setup in the docs.
LangGraph · CrewAI · ADK
Anything that speaks the OpenAI or Anthropic API works unchanged — point the base URL at Ainfera and set the model. Framework guides in the docs.