Keep control of your AI coding work.
Talk to the Conductor to understand your project and propose a change. Review the scope and diff, approve what gets applied, and run checks alongside the conversation. Maestro is a local desktop app that keeps the decisions, changes, and evidence in your hands.
v0.56.0 alpha. macOS Apple Silicon and Windows x64.
Where Maestro is heading
One conversation. You stay in control.
Maestro began by coordinating a creator, reviewers, and an approval loop. It is evolving into a desktop workspace centered on the Conductor: one place to discuss work, inspect your project, review proposed changes, and understand the results.
The principle stays the same: the agent proposes; Maestro enforces and executes; you hold final authority. Code, diffs, plans, checks, and evidence open alongside the conversation so you can inspect the work behind each decision.
Today, this experience supports project inspection, approved local edits, and named Checks. Future releases will bring more of planning, independent review, implementation, and closeout into the same flow.
Origin
Why Maestro exists
Maestro started when I started building a web app with a coding agent.
The first thing I asked for was a roadmap. The agent produced one full of Kubernetes, Redis, and Grafana — infrastructure for thousands of customers I didn't have.
Then I started implementing. Hardcoded values were everywhere, impossible to trace which one overrode which. The project crashed three times.
Writing development principles helped a little. But I never had time to truly review the plans, and during implementation the agent often improvised away from them.
What finally worked was using one agent to create plans and another to review them, iterating until they converged. The output was substantially more trustworthy. But I was spending most of my time copy-pasting between models.
So I built Maestro to automate the loop. Then I added a third agent — the Conductor — to summarize each round and point me at the decisions I actually need to make.
Today the Conductor is the center of the workspace: it helps you inspect the project, propose a bounded change, and understand the evidence. Independent creator/reviewer workflows remain available for larger planned work.
The goal is not to own the coding agent. The goal is to make agentic coding work reviewable, repeatable, and auditable.
One workspace
Keep the conversation and the evidence together
Ask a question, read the cited code, or review a proposed diff without leaving the conversation. Plans, findings, Checks, and logs open in a companion view when you need them.
- Inspect with context — The Conductor searches and reads project files, streams its answer, and links supporting evidence. A model's claim stays distinct from a read or result verified by Maestro.
- Decide with the diff in view — Approve a specific scope before a proposal is drafted. Review the file changes before Maestro applies them. Reject, revise, or roll back a change when needed.
- Stay involved — Follow live activity, steer a running inspection, or stop it. Project navigation, Roles, and Settings remain close at hand.
How it works
Start with a question. Make one small change.
Use a project you know and a result you can check for yourself.
- Open and inspect — Choose a local project and ask the Conductor to explain a file or behavior. Open the citations to check its answer.
- Describe the change — Name the intended result, the files it may touch, and what should stay unchanged.
- Approve the scope — Review the proposed file scope, allowed actions, and limits. This approval allows drafting a proposal; it does not change your source files.
- Review and apply — Inspect the proposed diff. Only a separate approval lets Maestro apply the patch and verify the resulting file contents.
- Check the result — Select an available Check, review any one-run risk approval, and inspect the logs and result. Complete the manual checks your change needs.
Available in v0.56.0
Know what Maestro can do today
Inspection, approved local edits, and named Checks are available now. Provider support and validation coverage have explicit limits.
What you can do
- Inspect projects with Claude. Read and search through the Conductor, with streamed answers, source citations, and bounded retrieval.
- Approve small multi-file edits. Review scope and diff separately; Maestro applies and verifies patches, with revision, rejection, and rollback controls.
- Run Checks on Windows and macOS. Select named validations, follow output, stop a run, and inspect saved logs and explicit success, failure, or indeterminate results.
- Use creator/reviewer workflows. Configure supported Claude Code, Codex, Gemini, or OpenCode tools for planned implementation and review. Multi-reviewer support applies to plan review.
- Keep a local record. Conversations, approvals, proposals, workflow artifacts, and validation evidence are stored with your project.
- Investigate failures. Inspect logs and local diagnostics, or export a source-free support packet.
What to expect
- The new inspection and editing flow currently uses Claude. Other tools supported by creator/reviewer workflows do not yet provide the same inspection and editing integration.
- Provider access is separate. Use your own authenticated account. The Claude inspection route requires an explicit confirmation that paid usage credits are disabled. Usage and cost are shown where available; unknown values remain unknown.
- Checks are selected by you. Current built-in coverage focuses on Cargo/Rust. Broader Rust, Node, and Python discovery is planned. Checks do not provide a terminal, deployment, or remote execution.
- Validation is not a full sandbox. A check may have filesystem or network effects outside Maestro's control. macOS process tracking is weaker than Windows Job ownership; uncertain cleanup is reported as indeterminate.
- Your judgment still matters. Reviewers can miss defects. Use the evidence and validate the result before shipping.
- The app runs locally; model requests reach your provider. There is no shared cloud workspace or organization administration. Selected project content is sent to the configured provider for model work.
Download
Download Maestro
v0.56.0 alpha. macOS Apple Silicon and Windows x64.
The macOS build is signed, notarized, and stapled. The Windows installer is unsigned. The Quick Start covers installation, Claude account setup, your first inspection, and approval of a small edit. See the Changelog for what changed since v0.43.0.
Community
Community
The Maestro community is on GitHub.
Discussions for ideas, feature requests, and workflow stories. Issues for bugs. I'm building this on my own, so replies may take a bit — everything gets read.