Prerequisites
API keys, model allowlists, gateway URL, and common setup issues.
Before configuring an SDK or agent, confirm these items in the Yaha console.
Project gateway key
Issue a yaha.* member key from Projects → API keys. Use the full token in Authorization: Bearer yaha.... or as the SDK api_key.
| Value | Meaning |
|---|---|
gateway.yahagateway.io | Production gateway hostname (no path) |
https://gateway.yahagateway.io/v1 | OpenAI-compatible base URL |
yaha.* | Your project member key |
Allowed models
Add models to the project plan allowlist. Requests for models not on the allowlist are rejected.
Upstream provider keys
Configure OpenAI and/or Anthropic provider keys on the organization or project so Yaha can forward requests upstream. The model you call must map to a provider you have configured.
How auth works
Your SDK/agent → Authorization: Bearer yaha.* → Yaha /v1/* → upstream LLMYaha verifies your yaha.* key, resolves the project, and swaps Authorization for the org's upstream provider key. You do not send provider API keys from the client.
Console wizard
In the console, open Integrations → Coding agents for copy-paste snippets per agent and project.
Troubleshooting
| Symptom | Fix |
|---|---|
403 forbidden | Invalid or revoked yaha.* key; check project membership |
model not allowed | Add the model to the project plan allowlist |
no provider credentials | Configure upstream provider key for the model's vendor |
| Claude Code hits api.anthropic.com | Set ANTHROPIC_BASE_URL to https://gateway.yahagateway.io (no /v1 suffix) |
| Codex ignores API key | Run /logout in Codex; set preferred_auth_method = "apikey" |
| Continue agent mode disabled | Add capabilities: [tool_use] to model config |
MCP 401 | Enable Tool Bridge on the project; use the same yaha.* key |