feed · watching · stats · help with your tokens
vercel-labs/agent-desktop
agent-desktop gives AI agents computer use on macOS, Windows and Linux by reading the OS accessibility tree instead of screenshots.
It lets an agent operate desktop applications through element references rather than pixel matching, while the control loop stays in the agent. A CLI and a C library share one core that reaches the operating system only through a platform adapter.
Similar
- storytold/craft-anything
Craft Anything is an open-source, pure Rust engine that studies software to understand its mechanics and uses AI agents to build new versions.
- KKKKhazix/agent-autoloop
Agent Autoloop lets a command-line agent pick up work in the background and follow it through, asking for a decision only when something is uncertain.
- Muvon/octomind
Octomind is an open-source coding agent harness that gives an LLM real access to a repository through MCP tools.
- govijr/terminal-mac
terminal-mac is a CLI tool that runs a separate macOS desktop in a virtual machine for computer-use agents, viewed and controlled from a terminal.
- jyotipmahes/ML-Interview-Prep
A repository containing Machine Learning Engineer interview preparation materials and from-scratch algorithm implementations.
- postmanlabs/postman-app-support
The public support repository where Postman users file feature requests, integration ideas and bug reports.
- JustVugg/colibri
colibri is a C inference engine that runs very large open mixture-of-experts models on the machine you already have.
- deepseek-ai/dsh-superagent-prototype
dsh-superagent is a research prototype for a proactive coding agent that works on a single GitHub repository using DeepSeek Harness.
sources
Claims below include README evidence. A matching quote or worker agreement does not establish that a claim is correct. Category majority among 2 workers. Version 1.
agent-desktop gives AI agents computer use on macOS, Windows and Linux by reading the OS accessibility tree instead of screenshots.
Computer use for AI agents: see and operate desktop apps on macOS, Windows, and Linux through the OS accessibility tree.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
It lets an agent operate desktop applications through element references rather than pixel matching, while the control loop stays in the agent.
agent-desktop gives an AI agent computer use without screenshots or pixel matching.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
A CLI and a C library share one core that reaches the operating system only through a platform adapter.
Both run the same core, which reaches the OS only through `PlatformAdapter`.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Reads the accessibility tree of a desktop app and returns JSON with a ref for every element the agent can act on.
It reads the accessibility tree of a desktop app and returns it as JSON, with a ref for every element the agent can act on.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
The agent clicks or types on a ref, reads the UI again and repeats.
The agent clicks or types on a ref, reads the UI again to check the result, and repeats.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
The loop lives in the agent; there is no model inside.
The loop lives in your agent; agent-desktop has no model inside.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Serves every command as an MCP tool over stdio.
`agent-desktop mcp` serves every command as an MCP tool over stdio.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Ships as both a CLI and a C library.
It ships as a CLI and a C library.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Actions are headless by default, with no pointer movement or focus change.
Actions are headless by default: no pointer movement, no focus change.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
A changed UI makes an action return STALE_REF or AMBIGUOUS_TARGET so the agent can take a new snapshot.
If the UI changed, an action returns `STALE_REF` or `AMBIGUOUS_TARGET`.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Carries built-in guides for agents that match the binary version.
The binary carries guides that match its version
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Has 60 command names, of which 56 are operational.
agent-desktop has 60 command names, 56 operational commands.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Ships as a CLI and a C library
It ships as a CLI and a C library
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Actions are headless by default, without moving the pointer or focus
Actions are headless by default: no pointer movement, no focus change
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
No model inside; the loop lives in your agent
The loop lives in your agent; agent-desktop has no model inside
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Changed UIs return STALE_REF so the agent takes a new snapshot
If the UI changed, an action returns `STALE_REF` or `AMBIGUOUS_TARGET`
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Snapshots, clipboard reads, screenshots and debug artifacts can contain private desktop content.
Snapshots, clipboard reads, screenshots, command output, and debug artifacts can contain private desktop content.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Needs macOS 13 or later, Intel or Apple Silicon.
macOS 13 or later, Intel or Apple Silicon
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Needs Windows 10 1809 or later, x64 or ARM64.
Windows 10 1809 or later, x64 or ARM64
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Needs Ubuntu 24.04 or later, x86_64 or aarch64, X11 or Wayland.
Ubuntu 24.04 or later, x86_64 or aarch64, X11 or Wayland
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Four key and mouse commands refuse to run until a daemon can release a held key or button.
The other four (`key-down`, `key-up`, `mouse-down`, `mouse-up`) refuse to run until a daemon can release a held key or button.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: There is no official Scoop package yet.
There is no official Scoop package yet.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Snapshots and screenshots can contain private desktop content
Snapshots, clipboard reads, screenshots, command output, and debug artifacts can contain private desktop content
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: No official Scoop package yet
There is no official Scoop package yet
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
ai-ml · developer-tools · cli