litellm-agent-control-plane

Why we included this project

Teams juggling several coding and general-purpose agents across different runtimes get a single web UI and REST API here, which removes the need to switch between separate consoles. You create agents, attach tools and skills, and choose the runtime per task: OpenCode, Claude Managed Agents, Cursor Agents API, Deep Agents, Hermes, or OpenClaw. The platform also handles the operational details that usually get in the way, with persistent sessions that survive across runs, CRON-based scheduling, shared memory so agents keep context, and access control that spares each developer their own cloud console credentials. A Docker Compose setup brings up the platform with Postgres and registers local runtimes automatically, so you can evaluate it on a single machine before wiring in hosted model providers.