Open Agent API

A small Go proxy that looks like the OpenAI Chat Completions API. Point Cursor BYOK (or any OpenAI SDK) at it, and it talks to the upstreams you already have logged in — no OpenAI sk- keys.

Surface	Auth	What it serves
Codex / ChatGPT	codex login → ~/.codex/auth.json	GPT-5.6 Sol/Terra/Luna, GPT-5.5, Spark
Gemini / Antigravity	Antigravity or Gemini CLI oauth	Gemini 2.5/3.x + Antigravity gateway Claude/GPT-OSS IDs
Claude Code	claude CLI login	Haiku / Sonnet / Opus / Fable

Default model is gpt-5.6-sol. Current release: v0.1.0.

Quick start

git clone https://github.com/teslashibe/open-agent-api.git
cd open-agent-api
codex login
docker compose up --build -d
curl -s http://127.0.0.1:8088/health
curl -s http://127.0.0.1:8088/v1/chat/completions \
  -H 'authorization: Bearer local-open-agent-api' \
  -H 'content-type: application/json' \
  -d '{"model":"gpt-5.6-terra","messages":[{"role":"user","content":"Say hi in five words."}]}' | jq .

Want the longer version? See Install. For Gemini 3.x, run scripts/sync-antigravity-auth.sh after login.

Where to go next

• Install — local run, Docker, ngrok
• Auth — Codex, Antigravity, Claude Code
• Cursor — BYOK settings and Agent mode
• Cursor tool conventions — how tool calls travel over the wire
• Model catalog — public IDs and good Cursor picks
• Contributing — tests, CI, PR habits
• Agents — short bootstrap for coding agents (AGENTS.md is the full one)

API at a glance

• GET /health
• GET /v1/models — filtered by GATEWAY_PROVIDERS (default codex,gemini,claude)
• POST /v1/chat/completions — streaming and non-streaming

JSON can be compressed (gzip / brotli / deflate) when the client sends Accept-Encoding. Streaming Agent turns stay uncompressed so SSE doesn’t get stuck in a buffer — details in API.

Model aliases live in internal/openai/models.go.