Long-running work stays in motion
Queue substantial work, continue across follow-ups, schedule what comes next, and return to a clear current state instead of supervising every turn.
Desktop-first operating system for AI-powered work
Keep ambitious work moving.
MAH is where people and AI agents run ongoing project work together. Set the goal and boundaries once; MAH coordinates execution, preserves context, checks progress, and returns the decisions that need you.
Research, analysis, writing, implementation, operations, and follow-up stay attached to one local project instead of scattered across chats and tools.
What MAH actually adds
A persistent work system—not another chat window.
MAH is not a code-completion tool, an outsourced delivery service, or a marketplace. It is the system that keeps your own work moving across agents, evidence, and human decisions.
Each step can use the right agent
Use different agents for research, planning, creation, verification, or recovery without rebuilding the project context for each handoff.
Progress stays reviewable
Keep the original outcome, task state, checks, and review evidence together so you can see the conclusion, the next action, and the details on demand.
Project work stays anchored locally
Your files, credentials, and project setup stay anchored on your Mac while MAH coordinates the agents and account resources around them.
Representative work
Built for work that crosses tools and takes more than one step.
For solo operators, builders, researchers, and teams whose AI work has outgrown a collection of chat windows.
Research and decide
Collect evidence, compare options, and prepare a decision while you keep ownership of the final call.
Create and improve
Move writing, implementation, documentation, review, and follow-up through the same project context.
Operate and follow through
Schedule recurring work, prepare releases, and let background tasks continue with a visible audit trail.
Agent integrations
Choose agents for their strengths. Keep the work in MAH.
Connect the agents you already use. MAH keeps the project, handoffs, review trail, and decisions together while each agent handles the steps it does best.
OpenAI Codex
Hands-on implementation, testing, review, and long changes inside a project.
Claude Code
Deep project understanding, research, careful planning, writing, and tradeoff analysis.
Cursor Agent
Useful for fast editor-native iteration when the work needs to stay close to the files.
GitHub Copilot
Fits GitHub-shaped work such as issues, branches, pull requests, docs, and cleanup tasks.
Google Antigravity
Adds Gemini-powered agent work when Antigravity is configured on the machine.
How it works
Humans set direction. Agents carry the work. MAH keeps the loop trustworthy.
01
Set the outcome and guardrails
Tell MAH what should be true when the work is done, what must stay within bounds, and which decisions remain yours.
02
Let MAH plan and route
MAH turns the outcome into reviewable steps and routes them across the agents already configured on your Mac.
03
Agents execute and check
Agents research, create, operate, and verify while MAH keeps progress, recovery, and evidence attached to the work.
04
Decide and keep moving
Review the conclusion and next action, make the decisions that need a person, and continue from the same project state.
Packages
Same MAH capabilities. Packages only change included resources.
Every package includes the same MAH capabilities. Purchases change seats, account machines, and monthly agent time only.
- Seats
- People covered by the account package.
- Account machines
- Desktop and Fleet machines that can join the account.
- Monthly agent time
- The account-level allowance for active agent work each month; new work waits when it is used up.
MAH is currently in private preview.