What DevPlane is. Unambiguously — for the people who'd have to commit to it.

DevPlane is a local-installed dashboard plus a CLI (dp) and an MCP server that coordinate assignments, agents, and completions across the full multi-tool agent ecosystem. One operator sits above a fleet of heterogeneous tools — Cursor, Claude Code, Codex, Replit, custom agents — and DevPlane is the layer between them: the board they all report to, and the instrument that records the coordination itself.

AI coding tools advertise productivity from the agent side — lines produced, tasks completed, time-to-PR. If the Ironies of Automation(Bainbridge, 1983) are operative — operator vigilance falling as agent reliability rises — those numbers systematically overstate the net effect, because the operator's real work is invisible to them.

So the distinction isn't “another agent dashboard.” It's the missing operator-side cockpit — and the difference is sharpest against the three things people reach for instead:

Inputs: assignments (cards), and the agents that run them, wherever they live. Method:a multi-tool kanban with claim semantics (an agent claims territory with a TTL, so parallel agents don't collide), a two-phase actor handoff (builder ships → reviewer audits, and the reviewer's transition requires an artifact the builder can't produce), and a completion-block protocol. Outputs: a coordinated dispatch and a continuous, queryable coordination-event log.

• —Dispatch a cloud agent (e.g. Cursor) straight from a card and track the run to a PR.
• —Run several agents in parallel without collisions — claims arbitrate contention.
• —Close every assignment with a typed completion block, not a free-text note.
• —Query the portfolio: cross-project status, a pattern/capability library, handoffs.
• —Reach the same primitives from the terminal (dp) or from an agent (MCP tools).
• —Produce research-grade telemetry as a by-product of doing the work.

The load-bearing artifact behind all of it is the completion block. Here is one, shown end to end rather than described:

{
  "asn": "DP-176",
  "agent_id": "cursor-cloud-7f3a",
  "status": "done",
  "sha": "173b702",
  "summary": "Aligned the 'two surfaces, one product' framing across internal docs.",
  "follow_ups": ["devplane-site /explain section (Explainer Standard)"],
  "blockers": [],
  "coordination_events": 14
}

DevPlane is the engineer-facing half of a “two surfaces, one product” story. devplane.dev speaks to engineers; Performix /ai speaks to business operators weighing AI-transformation readiness. They are two faces of the same underlying capability — the shared data-substrate / AI-ETL layer that prepares real organizational data for use.

One boundary, stated plainly so it isn't over-read: Performix's diagnostic engine is psychometrics, not an LLM — AI is the consumer doing substrate preparation, not the thing doing the measuring. DevPlane and Performix share the framing and the substrate; they do not share internals beyond that. See the operator surface →

Where it's going. DevPlane is also the apparatus for a pre-registered field study (C1) of risk compensation in human-AI coordination — the lead arm of a research program on coordination cost. The instrument and the product are the same surface: as the tool earns a wider installer base, the study earns a larger sample. That dual role is the point, not a side effect. Read the research program →