GET /v1/models; this docs site is updated
periodically and GET /v1/models is always the source of truth.
Point the OpenAI SDK at https://api.meshrouter.app/v1, authenticate with a mesh_live_... key,
and call the API the way you already do. Or open the MESH chat to
talk to any model in your browser, no SDK setup required.
One endpoint, every model
Use router modes (
auto, fast, cheapest, best) or address a specific provider model.Pay per call in USDC
Top up with USDC on Base or Solana. Calls draw against the balance; no subscription, no
permanent free tier.
Receipts on every call
Response headers carry upstream cost, MESH margin, total USD, and a receipt ID for lookup.
Per-key limits
Per-key request, token, and monthly USD ceilings. Revoked keys stop working in the same request.
How it works
- Get an API key starting with
mesh_live_. - Set
OPENAI_BASE_URL=https://api.meshrouter.app/v1. - Call the API the way you’d call OpenAI.
- Read the receipt headers off the response.
Install an SDK (optional)
| Language | Package | Install |
|---|---|---|
| TypeScript / JavaScript | @meshrouter/sdk | npm install @meshrouter/sdk |
| Python | meshrouter | pip install meshrouter |
| Anything OpenAI-compatible | use the official OpenAI SDK | swap OPENAI_BASE_URL |
Next steps
- Open MESH chat — no API key needed, talk to any model in the browser.
- Quickstart — two-minute first request.
- Models — full catalogue (80+ models across every major family); call
GET /v1/modelsfor the live source of truth. - Tool calling, structured output, vision.
- Authentication — keys, limits, revocation.
- Accounts — project structure and credit balance.
- Provider health — live model + provider status.
- TypeScript SDK (
@meshrouter/sdk) and Python SDK (meshrouter). - Privacy classes — pick one of four per-call classes (Standard, Protected, Private Gateway, Confidential). The router enforces the class you ask for and refuses to silently downgrade.
- Tokenomics — 100% of MESH margin flows back into $BETTER through buyback, lock, distribute, or burn.