Pre-1.0 · research preview · self-installed
An operator cockpit for human-AI software development. The measurement layer that agent-side productivity metrics miss.
A multi-tool kanban with two-phase actor handoff, completion-block protocol, and continuous coordination-event log. Built for the operator running Cursor + Claude Code + Replit + custom agents concurrently on a real codebase. Used today by its author; the instrument behind a pre-registered field study of risk compensation in human-AI coordination.
The problem
The productivity story is being told from the wrong side.
AI coding tools' productivity claims rest on agent-side measurements — lines produced, tasks completed, time-to-PR. If the Ironies of Automation (Bainbridge 1983) are operative — operator vigilance falling as agent reliability rises — those measurements systematically overstate net effect.
No operator-side cockpit catches the loss. The operator running multiple agents on a real codebase is doing coordination work that's invisible in any single tool's dashboard. That work is load-bearing. It is also under-measured to the point of being under-considered.
What DevPlane is
The cockpit. The instrument. The protocol.
DevPlane is a local-installed, browser-based dashboard plus a CLI (dp) that coordinates assignments, agents, and completions across the full multi-tool ecosystem. The same surface that runs daily operations also produces the coordination-event log that grounds the C1 risk-compensation field study.
Multi-tool kanban
Cards for every assignment, regardless of which tool the agent runs in. Cursor, Claude Code, Codex, Replit, custom — coordinated through one board, not four dashboards.
Two-phase actor handoff
Builder ships → reviewer audits. The reviewer's transition requires an artifact the builder can't produce. Enforces review without trusting it.
Completion-block protocol
Every assignment ends with structured machine-readable completion, not free-text close-out. Status, SHA, summary, follow-ups, blockers — typed.
Coordination-event log
Continuous production telemetry on a real operator running real agents on a real codebase. Research instrument, not just an audit trail.
The research
Built and instrumented to falsify a specific prediction.
DevPlane is also the apparatus for a pre-registered empirical program on coordination cost in heterogeneous AI-tool ecosystems. The lead study (C1) tests whether risk compensation — operator vigilance falling as agent reliability rises — is operative in this regime.
The methodology generalizes beyond AI agents. Any team running heterogeneous tools through a coordination layer (multi-tool ops dashboards, hospital handoff systems, distributed scientific instruments) shares the same shape of problem.
Status
Honest about where this is.
Pre-1.0 · research preview
DevPlane is currently a local-installed tool used daily by its author. The CLI ships; the dashboard runs at localhost:4000. There is no hosted product yet. The pattern is being stabilized in production-of-one before being opened to a broader installer base.
If you're a solo founder, lead engineer, or research-leaning practitioner running multiple AI coding agents and would consider being an early external operator — the C1 study needs exactly this — the fastest path is a short note.
Request early access →