Enables Claude Code to delegate coding jobs to cheaper or free models through opencode while staying the sole interface, relaying permission requests and questions back to the user and running multiple jobs in parallel git worktrees with diff review, verification, cost caps, and merge checks.
Enables Claude Code and other MCP clients to offload bounded work — code review, boilerplate, commit messages, format conversion, embeddings — to a local or cloud OpenAI-compatible LLM, cutting token spend by keeping large files out of the orchestrator's context. It also provides a second model for reviewing code and drafting tasks while the main assistant retains planning and QA.