Plan, execute, verify, and review code changes autonomously. LangGraph-powered multi-phase pipeline with human-in-the-loop oversight.
Open Workspace →Every instruction flows through a deterministic graph. Each phase has its own model, tools, and validation gates.
Not a toy. Shipyard Agent handles multi-file edits, test suites, type systems, and CI gates.
Watch file edits stream in real-time as the agent works. Every tool call, every diff, visible in the dashboard as it happens.
Review the agent's plan before any code is touched. Approve, modify, or reject. Like Cursor's plan review, built into the pipeline.
Switch between GPT-5.4 Mini for fast edits and GPT-5.4 for deeper reasoning. Per-run model selection.
Capture snapshots of type safety, test health, security, build speed, and more. Compare original vs refactored codebases with radar + trend charts.
A TypeScript runtime with Express API, WebSocket streaming, and Postgres persistence. The graph engine is LangGraph.
LangGraph state machine with plan, execute, verify, review, report nodes. Repo map context, tool hooks, and persistence layer.