Instead of having the client harness integrated to the model steaming logic, could you implement mac-router as just a valid OpenAI steaming endpoint / proxy with your SSD logic?
We could then use it with any existing and future agent (codex, opencode, cursor, openclaw etc.)
Instead of having the client harness integrated to the model steaming logic, could you implement mac-router as just a valid OpenAI steaming endpoint / proxy with your SSD logic?
We could then use it with any existing and future agent (codex, opencode, cursor, openclaw etc.)