PlanWright is the control plane for agentic engineering that turns chaotic inputs—transcripts, decks, email, Slack—into structured objectives and triages acceptance so humans only rule on what needs judgment. PlanWright inverts the two human‑paced ceremonies: planning becomes chaos‑in, objectives‑out, and acceptance becomes triage‑don’t‑drown, letting agents work at machine speed while every ruling is signed. Using MCP, the system logs each step in a hash‑chained, ECDSA P‑256‑signed audit trail that satisfies SOC 2 change‑approval controls. Agents claim objectives via Claude Desktop, execute, request acceptance, and the machine clears mechanical criteria, routing only critical judgments to humans for sign‑off.

The audit record is immutable and verifiable on a public trust page. Pricing ranges from free for solo developers to enterprise tiers with multi‑year retention, SSO, and custom integrations. Getting started requires a single command that opens a browser for GitHub login, after which the connection is cached for future sessions.

PlanWright ships the MCP server, audit chain, and board as core primitives, enabling teams to scale agent‑native planning while maintaining compliance and traceability.