Tools
Coding assistants, chat clients and SDKs only need two settings changed: the Base URL and the API Key. The command-line tools below were all run successfully with the configuration in this documentation.
Prefer not to edit config files?
Use our desktop app CodingRouter Connect: sign in, pick a token and a model, and Claude Code, Codex and Gemini CLI are set up for you.
| Tool | API used | Base URL | Token group | Recommended models |
|---|---|---|---|---|
| Claude Code | Claude Messages | https://api.codingrouter.ai | claude | claude-sonnet-5-5, claude-opus-5-5 |
| Codex CLI | Responses | https://api.codingrouter.ai/v1 | codex_gpt or gpt | gpt-6.1-sol, gpt-5.5 |
| Gemini CLI | Gemini | https://api.codingrouter.ai | gemini | gemini-3.5-flash, gemini-3.7-flash |
| OpenCode | Chat Completions | https://api.codingrouter.ai/v1 | Depends on the model | Any text model |
| Chat clients and SDKs | Depends on the client type | See the page | Depends on the model |
The token's group must contain the model the tool uses. A wrong model name, or a model not in the group, returns 503. See Errors and retries.
Use a separate token per tool
Create a separate token for each tool and set a quota limit. Usage can be tracked separately, and if a token leaks you only need to disable that one.
