Connect Your Coding Agent in Minutes
Point Cursor, Claude Code, Codex CLI, and other agents at Yaha with one endpoint and a project key.
AI coding agents are only as good as the gateway in front of them. When every developer configures their own provider key, you get shadow spend, inconsistent model policy, and zero audit trail.
Yaha gives agents a single OpenAI-compatible endpoint and a project-scoped key, so Claude Code in one repo and Cursor on a laptop both route through the same quotas, allowlists, and logs.
One URL for every agent
The gateway URL is:
https://gateway.yahagateway.io/v1
Use your yaha.* project key as the API key. Agents that support custom OpenAI base URLs work without plugins or forks.
Claude Code
Set the base URL and API key in your Claude Code configuration to point at Yaha instead of a provider directly. Full step-by-step instructions live in our Claude Code integration guide.
Cursor and other IDEs
Any tool that accepts a custom OpenAI-compatible endpoint follows the same pattern: base URL → gateway, API key → Yaha project key. Swap models through the console allowlist instead of editing each developer machine.
Codex CLI and Continue
We maintain guides for Codex CLI and Continue with copy-paste config snippets. You can mix providers behind one key.
Why route agents through Yaha
- Spend caps per project stop runaway agent loops
- Model allowlists prevent accidental calls to premium models
- Central logs show which repo or user drove which requests
- Smart routing and caching apply to agent traffic, not just backend services
Agents generate bursty, repetitive prompts, exactly the traffic profile caching and routing are built for.
Quick verification
After configuring your agent, send a test completion:
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": "Say hello from my coding agent setup"}]
}'
Check Projects → Logs in the console. You should see the request within seconds.
Connecting an agent takes minutes; governing agent spend without a gateway takes quarters of invoice surprises. Pick the faster path.
Create a project key: Sign in to the console and follow the getting started docs.