API Reference Overview
Base URL
https://api.openproxy.ai
Two API surfaces
Proxy endpoints — OpenAI-compatible, used to actually call a model:
POST /v1/chat/completions
POST /v1/embeddings
GET /v1/models
Management endpoints — used to configure your organization, under /api/v1/:
/api/v1/auth/*
/api/v1/api-keys
/api/v1/provider-keys
/api/v1/organizations/current
/api/v1/teams
/api/v1/analytics/*
/api/v1/prompt-templates
/api/v1/experiments
Authentication
Both surfaces use a bearer token, but different tokens: proxy endpoints take an API key (Authorization: Bearer sk_live_... or opai_... depending on how it was issued), management endpoints take a session token issued by /api/v1/auth/login. See Authentication.
Response headers on proxy requests
Every proxy response includes gateway metadata as response headers, regardless of which client you use to call it:
| Header | Meaning |
|---|---|
x-openproxyai-request-id | Unique ID for this request, for correlating with the audit log |
x-openproxyai-provider | Upstream provider actually used |
x-openproxyai-model | Model actually used |
x-openproxyai-cost-usd | Cost of this request in USD |
x-openproxyai-latency-ms | Total latency including proxy overhead |
x-openproxyai-ttft-ms | Time to first token (streaming requests) |
x-openproxyai-cache | Which cache tier served this response, if any |
x-openproxyai-policy-action | Policy action taken, if any (e.g. redacted) |
x-openproxyai-policy-reason | Human-readable reason for the policy action |