Overview
Gateway URL, project key, supported endpoints, and your first request.
Yaha AI Gateway is an OpenAI-compatible gateway for applications, SDKs, and coding agents. Point your client at the gateway URL and authenticate with a yaha.* project key. Your application does not need provider API keys.
1. Get a project API key
- Sign in to the console.
- Open your project → API keys.
- Create a member key. It starts with
yaha.; use the full string as your API key.
See Prerequisites for allowlists, upstream keys, and troubleshooting.
2. Set the gateway URL
Use this gateway base URL. Include /v1 for OpenAI-compatible clients:
https://gateway.yahagateway.io/v1Most OpenAI-compatible clients accept a base URL ending in /v1. Some Anthropic-compatible tools use the host only without /v1; see the Claude Code guide.
3. Pick a model
List the models available to your project:
curl https://gateway.yahagateway.io/v1/models \
-H "Authorization: Bearer yaha.YOUR_PROJECT_KEY"Use one of the returned model IDs in your request body. Model access is controlled by your project configuration.
4. Send a test request
curl https://gateway.yahagateway.io/v1/chat/completions \
-H "Authorization: Bearer yaha.YOUR_PROJECT_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-4o-mini",
"messages": [{"role": "user", "content": "Hello from Yaha"}]
}'Replace yaha.YOUR_PROJECT_KEY with your key and pick a model on your project allowlist.
Supported public endpoints
The gateway accepts common OpenAI-compatible /v1/* routes, including:
- Text:
/v1/chat/completions,/v1/responses,/v1/messages,/v1/completions - Embeddings and ranking:
/v1/embeddings,/v1/rerank - Image, audio, and video:
/v1/images/generations,/v1/images/edits,/v1/audio/speech,/v1/audio/transcriptions,/v1/videos - Moderation:
/v1/moderations - Discovery and reporting:
/v1/models,/v1/generation?id=REQUEST_ID
Streaming is supported for clients that set "stream": true on compatible generation endpoints.
Next steps
- Language examples: examples for common backend, frontend, mobile, and scripting languages
- Agent integrations: Claude Code, Codex, Cline, Continue, Hermes, MCP