v0.1.3 — Now with Anthropic & OpenAI compatible APIs

One API for every AI provider

Self-hosted gateway that unifies Claude, OpenAI, Gemini & Ollama behind OpenAI and Anthropic compatible APIs. Use your own subscriptions. Works with any SDK. Zero external dependencies.

Get StartedSee how it works
Terminal
$ curl -fsSL https://raw.githubusercontent.com/
  openhinge/openhinge/main/install.sh | bash
$ cd ~/openhinge && npm start

  Open the dashboard to set your password
  Dashboard: http://localhost:3700

Every major provider. One endpoint.

Use OAuth with your existing subscriptions or bring your own API keys.

C
Claude
Anthropic
Multi-account OAuth, keychain import, or API key
O
OpenAI
ChatGPT / Codex
Codex CLI auth or API key
G
Gemini
Google
Google OAuth or API key
O
Ollama
Local
Auto-detected on localhost

Everything you need to run AI in production

Not just a proxy. A complete gateway with auth, routing, logging, and a dashboard.

OpenAI + Anthropic APIs
Both /v1/chat/completions and /v1/messages endpoints. Works with OpenAI SDK, Anthropic SDK, or curl. Zero code changes.
Soul-Based Routing
AI personas with dedicated system prompts, models, and endpoints. Route by URL, header, or key binding.
OAuth Login
Use your Claude Pro, ChatGPT Plus, or Gemini subscription. No per-token cost. Tokens auto-refresh.
API Key Management
Generate ohk_ keys with rate limits, soul binding, budget controls, and expiry. Revoke instantly.
Full Streaming
Server-Sent Events in both OpenAI and Anthropic formats. Stream responses from any provider with consistent output.
Request Logging
Every request logged with tokens, latency, and cost. Search, filter, sort, and paginate.
Fallback Routing
Providers have priority. If one fails or goes down, the next healthy provider handles the request.
Web Dashboard
Manage providers, souls, keys, logs, and settings from a modern dark-mode UI. No CLI needed.
Zero Dependencies
Embedded SQLite. No Redis, Postgres, or Docker required. One directory, one command to run.
OpenClaw Integration
Generate ready-to-paste config for OpenClaw. One-click key creation with auto-configured provider block.
Multi-Account Providers
Add unlimited accounts of the same provider. Stack 10 Claude subscriptions with automatic fallback between them.
Cloudflare Tunnel
Expose your gateway over HTTPS with Cloudflare Tunnel. No open ports, no firewall config. Connect from anywhere.

Works with everything

Drop-in replacement for both OpenAI and Anthropic. Native integration with developer tools.

OpenAI SDK
Python, Node.js, Go, Rust — any library that speaks the OpenAI protocol works out of the box.
base_url = "http://localhost:3700/v1"
Anthropic SDK
Full /v1/messages support with streaming. Point the Anthropic SDK at your gateway with one line.
base_url = "http://localhost:3700"
OpenClaw
Generate a key from the dashboard and get a ready-to-paste config block for openclaw.json.
"primary": "openhinge/claude-sonnet-4-6"

Three commands. That's it.

Install, start, and you're live. No configuration files to write.

1
Install
curl -fsSL https://openhinge.com/install.sh | bash
2
Start
cd ~/openhinge && npm start
3
Use
curl http://localhost:3700/v1/chat/completions \
  -H "Authorization: Bearer ohk_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-sonnet-4-6","messages":[{"role":"user","content":"Hello"}]}'

On first start, OpenHinge generates your config automatically.
Open http://localhost:3700 and set a password to access the dashboard.

How requests flow

Every request passes through auth, rate limiting, soul routing, and provider selection.

architecture
  Client Request
      │
      ├── Auth (validate ohk_ key)
      ├── Rate limiter (sliding window, per key)
      ├── Budget check (daily/monthly limits)
      │
      ├── Soul resolver (system prompt + model)
      ├── Provider selector (priority + health + fallback)
      │
      ├── Provider adapter
      │   ├── Claude   (OAuth or API key)
      │   ├── OpenAI   (OAuth or API key)
      │   ├── Gemini   (OAuth or API key)
      │   └── Ollama   (local)
      │
      ├── Stream / collect response
      └── Log usage (tokens, latency, cost)

Zero token cost. Use your subscription.

Stop paying per-token API fees. OpenHinge uses your existing Claude Pro, ChatGPT Plus, or Gemini subscription — the same plans you already pay for.

No API billing
OAuth login uses your flat-rate subscription. No per-token charges, no surprise bills, no usage caps beyond your plan.
Tunnel & expose
Add a Cloudflare Tunnel and your gateway gets a public HTTPS URL. Connect your apps, bots, and services from anywhere.
Your infra, your rules
Self-hosted on your machine. Your data never passes through third-party proxies. Full control over access, rate limits, and logging.

How it works: Install OpenHinge on your Mac or server. Add your Claude/OpenAI/Gemini subscription via OAuth. Set up a Cloudflare Tunnel for a public URL. Point your apps at the tunnel URL — they talk OpenAI protocol, OpenHinge translates to any provider. Your subscription covers the cost.

Stop paying per-token API fees

One gateway. OpenAI + Anthropic APIs. Every provider. Use your existing subscriptions at zero extra cost. Self-hosted in under a minute.

View on GitHub