Breaking

Maestro Launches as Open-Source Structured Workflow Layer for AI Coding Agents With 25 Commands

June 30, 2026 at 14:14 EDT

AI Agents
Software Dev & Coding
Open Source

Developer Tom Dörr has released Maestro, an open-source project that bundles structured workflows for AI coding agents. It combines a single core skill, 25 commands, seven domain references, a persistent memory layer and an audit trail to bring discipline to how agents operate.

Open Source · MIT License

Maestro adds a workflow layer to AI coding agents

A single core skill plus 25 commands, a persistent memory layer that survives across sessions, and a full audit trail — retrofitted onto Cursor, Claude Code and 10+ other tools.

commands across 4 categories

10+

supported AI dev tools

domain reference files

±20%

cost-estimate accuracy in audit log

The 25 commands, by category

Analysis

/diagnose · /evaluate · /reflect

Fix & Improve

/refine · /streamline · /fortify · /zero-defect

Enhancement

/amplify · /compose · /chain · /turbocharge

Utility

/extract-pattern · /onboard-agent · /capture · /recap

A self-reported productivity claim

Developer report — used as a strict control layer for Claude Code

0 hrs

saved without it

15 hrs / week

cut from debugging loops

Broader benchmarks are still limited — real-world evaluation remains early.

Persistent memory in a .maestro/ directory

Git-friendly, opt-in state that survives once a session ends — added in v2.

context.md

working context

decisions.jsonl

decision log

audit.jsonl

duration · tokens · cost

sessions/

per-session state

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.