Connect AI tools to RouterBrain with an API key from the console and the shared model catalog—no per-upstream key juggling.
Prerequisites
- Create an inference API key in the console (API keys).
- Pick a model id from the catalog (List models).
- Choose OpenAI-compatible or Anthropic-compatible base URL per tool (Capabilities).
| Surface | Base URL |
|---|---|
| OpenAI-compatible | https://51kik.com/v1 |
| Anthropic-compatible | https://51kik.com/anthropic (client base; API path /v1/messages) |
Tool guides
| Tool | Recommended surface | Guide |
|---|---|---|
| Cursor | OpenAI-compatible | Cursor |
| OpenClaw | OpenAI-compatible | OpenClaw |
| Claude Code | Anthropic-compatible | Claude Code |
| TRAE | OpenAI-compatible | TRAE |
| Hermes Agent | OpenAI-compatible | Hermes Agent |
| Codex | OpenAI-compatible | Codex |
| Other tools | Depends on client | Other tools |
Troubleshooting
- 401 / 403: verify key and model access (Errors).
- Unknown model: use catalog
id, not hard-coded vendor model strings. - Attribution: optional
x-trace-id,x-user-id,x-agent-name(Request correlation).