YOUR LOCAL AGENT WORKSHOP Public early release forthcoming

One backlog.
Any agent.
You conduct.

Local-first planning and orchestration for one developer. Turn a Markdown backlog into reviewed plans and verified merges.

Local-first · One developer · Your repositories

MEET PLANESTRO

A local tool for a solo developer coordinating coding agents. Claude plans and orchestrates. Claude or Codex implements. Planestro manages the steps between them, with your backlog, plans, and decisions saved in Markdown.

THE WORK, IN VIEW

Your next decision.
Not another terminal to watch.

See what’s running, what’s ready, and what needs your judgment. Open the plan, ask for a change, or let approved work move forward.

PLANESTRO / PLAN REVIEW LOCAL WEB APP
A plan ready for a decision · Historical product captureOpen full screenshot ↗

01 / PARALLEL WORK

A worktree for every item.

Agents work on separate branches. Dependencies keep related work in order; epics bring it together on an integration branch.

02 / DURABLE STATE

Your files are the memory.

Plans, notes, and activity live in a Git-versioned data directory. Decisions survive restarts and replacement sessions.

03 / HUMAN JUDGMENT

Answer when you’re ready.

A question parks the session and saves the request. Return to a plan, permission, or review when it has your attention.

04 / CONTROL

Try it. Pause it. Take over.

Start with a dry run. Pause new automation, hold a PR for review, or finish an item yourself. Cleanup requires evidence.

A SMALL FIRST STEP

Your project.
Your machine.

Install Planestro, connect a new repository, and open your local board. Start in dry-run before turning on real agents.

Follow the four-step guide
  1. 01 Prepare your tools
  2. 02 Install Planestro
  3. 03 Connect a project
  4. 04 Open your board

Node 24 + Git + GitHub CLI + Claude Code (signed in).
Codex implementation is optional.

Just looking? Try the no-login sample →

FIELD NOTES / 001

Built while
building something real.

From the workshop →
Hurricane Harrison explores a lantern-lit town with a slime and a skeleton

BUILDING WITH AN AGENT FLEET

How I built an AI fleet and shipped Hurricane Harrison to the App Store.

From a personal playbook to a working orchestration system. The architecture, quality checks, and hard limits along the way.

One operator.
A whole fleet.

For a web app, an API, a game, or any project with Git and CI. Run your own instance, use your own credentials, and stay accountable for what ships.

Get to know the workflow →