BuildDevelopers

Your gateway, at a glance.

Scoped API keys with their own spend caps and rate limits, and live requests, tokens, latency and credit burn for every one of them.

Documentation

OpenAI-compatible base URL: https://infermesh.dev/v1
Anthropic SDK base URL: https://infermesh.dev
Auth header: Bearer sk-infer-…
Protocols: Chat Completions, Messages, MCP

Quickstart

Point your SDK at InferMesh

Docs

OpenAI-compatible base URL

https://infermesh.dev/v1

Anthropic SDK base URL

https://infermesh.dev

curl https://infermesh.dev/v1/chat/completions \
  -H "Authorization: Bearer sk-infer-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-sonnet-5.5",
    "messages": [{ "role": "user", "content": "Explain proof of history in one sentence." }]
  }'

Documentation

Six pages cover the gateway end to end.

What a key carries

Every key is scoped, capped and rate-limited on its own.

Scopes: inference (chat completions and messages), account (read balance and usage), keys (create child keys) and mcp (agent tools).
Spend caps: An optional ceiling in credits per UTC day, per UTC month or for the key's lifetime. In-flight requests count, so concurrent ones can't overshoot it.
Rate limits: A requests-per-minute limit on each key: 600 by default, up to 10,000.
Storage: A key is shown once. Only a salted hash is stored, so a lost key is revoked and replaced, never recovered.

Building something that runs on its own?

Give an agent an isolated balance, a scoped key, MCP tools and a USDC vault it can draw on only within limits you sign on-chain.

Set up an agent MCP guide