# PlanWright

> The control plane for autonomous software labor.

PlanWright is the control plane for autonomous software labor. Humans declare objectives — the desired state — with no story points, velocity, or capacity math. Coding agents (Claude Code, Claude Desktop, Cursor, Codex, Amp) claim those objectives over a native Model Context Protocol server, decompose them, execute, and check in; claims lock so parallel agents don't collide. Every state change — human or agent — is written to an append-only, hash-chained ledger signed with per-workspace ECDSA P-256 keys in AWS KMS: tamper-evident audit evidence built toward the 2026 AICPA SOC 2 Trust Services Criteria for AI-generated code, ISO 42001, and the EU AI Act.

It looks like a board. It's actually a control plane.

## Features

- Objective-native planning (post-kanban) — objectives are atomic; no story points, velocity, or capacity math
- Collision-free parallel agents — claiming an objective locks it so other agents skip it
- Bugs live in a separate lane, never mixed into the planning surface
- Native MCP server (Streamable HTTP, OAuth 2.1 + PKCE + Dynamic Client Registration) — Claude Code, Claude Desktop, Cursor, Codex, and Amp connect directly; service tokens for CI/headless
- Immutable, append-only, hash-chained audit ledger with per-workspace ECDSA P-256 signing keys in AWS KMS
- Enforced human acceptance gates — agents can request acceptance but cannot mark their own work done
- Public trust page per workspace with independent audit-chain verification
- GitHub OAuth login and GitHub App for repo integration
- Cloud SaaS at https://planwright.tools; self-hosted option on roadmap

## Installation & distribution

- MCP server: `claude mcp add planwright --url https://mcp.planwright.tools/mcp` (Streamable HTTP; legacy SSE at /sse deprecated, sunset 2026-10-01, returns 410)
- Published on the official MCP Registry as `tools.planwright/planwright`
- Agent skill on ClawHub / OpenClaw: `openclaw skills install @toddamerrill/planwright-control-plane`
- Agent skill on Hermes: `hermes skills install Planwright/planwright/skills/planwright-control-plane`
- Source-visible skill: https://github.com/Planwright/planwright

## Target Audience

- Indie hackers and agentic builders running parallel coding agents (throughput, collision-free fan-out)
- Small to mid-market engineering teams (2–100 developers) standardizing on agentic coding tools
- Regulated industries (fintech, healthtech, govtech) needing audit trails for AI-generated code
- Enterprise buyers requiring SOC 2 Type II evidence bundles for AI development workflows

## Use Cases

- Give coding agents a first-class control plane to claim work against and check in to
- Run 3–5 parallel Claude Code / Cursor / Codex sessions without collisions
- Produce auditor-ready, tamper-evident compliance evidence for AI-authored code
- Track objective-level outcomes instead of task-level breakdown theater
- Enforce human acceptance gates on every AI-generated change
- Replace Jira and Linear with an agent-native planning surface

## About

Built by Todd Merrill / Silverback CTO. PlanWright is a Silverback venture targeting the gap between traditional project-management tools and agent-native coding workflows.

Contact: todd@silverbackcto.com
Booking: https://fantastical.app/toddmerrill/silverback-cto

## Pages

- [Home](https://planwright.tools): Product overview and positioning
- [For Regulated Teams](https://planwright.tools/regulated): HIPAA, SOC 2, FedRAMP, FINRA positioning
- [Compliance Brief](https://planwright.tools/compliance): Technical brief for compliance teams and external auditors
- [Compare](https://planwright.tools/compare): PlanWright vs Linear, Jira, Plane
- [Pricing](https://planwright.tools/#pricing): Free, Team, Business, Enterprise tiers
- [MCP Server Setup](https://planwright.tools/docs/mcp-setup)
- [Connect a Coding Agent](https://planwright.tools/docs/how-to/connect-coding-agent)
- [Install as an Agent Skill (ClawHub / Hermes)](https://planwright.tools/docs/how-to/install-as-skill)
- [Docs Home](https://planwright.tools/docs)

## Technical

- Stack: Next.js 14 (App Router) + TypeScript, PostgreSQL 16 on RDS with row-level security, AWS ECS, Claude Sonnet via Bedrock
- MCP Server: Model Context Protocol over Streamable HTTP with OAuth 2.1 + PKCE + Dynamic Client Registration (RFC 7591); legacy SSE transport deprecated (sunset 2026-10-01, returns 410)
- Audit signing: ECDSA P-256 via AWS KMS
- Auth: GitHub OAuth for human login, OAuth-issued JWT for MCP, optional pre-shared MCP tokens for CI
- Launch: 2026