Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

zero-use-computer

Rust MCP server for desktop control via accessibility trees, with an optional decision model (Jev or compatible) for small choices.

MCP serverproduction
View on GitHub ↗

Overview

Zero Use Computer gives AI agents desktop control by reading apps the way a screen reader does. Instead of sending a screenshot every turn, it gives the model the window's accessibility tree with numbered elements, so click(12) presses a real control through the operating system's accessibility API, often on background windows without moving the user's mouse. It looks at pixels only when they matter. It is one Rust binary for any MCP client on Windows, macOS and Linux, and updates itself.

It remembers screens so a returning view only sends what changed, and each action reports its own result and whether an expectation was confirmed. According to the list summary and README, an optional decision model, Jev or any compatible server, can be handed the small choices so the main agent does not spend turns on them. Benchmarks and a guide are included.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich numbered element should be acted on for this small step?(inferred)—

How it works

  1. The agent requests app state and gets a numbered accessibility tree.
  2. It batches actions such as set, click and select.
  3. Optionally, a decision model like Jev handles small choices.
  4. Each action returns its resulting state.

Strengths

  • Far fewer screenshots and tokens.
  • Works on background windows.
  • Cross-platform single binary.

Caveats

  • Jev is optional and a small part of the system.
  • Depends on apps exposing good accessibility data.

Getting started

Download from https://github.com/mhrsdev/zero-use-computer/releases/latest and connect an MCP client.

At a glance

Type
MCP server
Status
production
Language
Rust
Runtime
MCP server (Windows, macOS, Linux)
Has code
yes

Use cases and integrations

How curated lists describe it

  • Computer use over MCP that reads what a screen reader reads and acts on real controls, with an optional decision model - Jev or any compatible server - for the small choices.

    MrJev/awesome-jev

Frequently asked questions

Does it use screenshots?

Only when pixels matter; it mainly uses accessibility trees.

What role does Jev play?

An optional decision model for small choices.

Which platforms?

Windows, macOS and Linux.

Related projects

  • agent-desktop

    Rust CLI for desktop computer use via accessibility trees, with optional jev-desktop scripts letting Jev pick the control and action.

    Agent skillbetaRust
  • arc-cua

    macOS computer-use package: a background app driver for any agent, plus a loop where Jev picks each UI action for a bounded subtask.

    ProjectbetaPython
  • Flick

    Local stdio MCP server that executes whole browser and macOS goals with a Jev-driven observe-decide-act loop.

    MCP serverbetaTypeScript
  • jev-adb

    MCP server that lets agents observe and control Android phones over ADB, with Jev picking controls and judging goal completion.

    MCP serverbetaTypeScript
  • jev-browser-skill (ChenYCL)

    Browser and computer use for coding agents via CLI and MCP, with Jev's calibrated judgments and a code-owned control loop.

    MCP serverbetaTypeScript
  • Jev-Desktop

    Windows MCP server and CLI where Jev observes windows and chooses controls so agents can hand off whole desktop tasks.

    MCP serverbetaPython