Clawd Cursor is free software that lets AI agents control desktop apps for GUI tasks where a native API or CLI is unavailable. It builds a confidence-scored screen map from accessibility data and OCR, assigning stable element IDs and using screenshots only when needed. Agents can click and type, and actions can be checked against an expected result; each tool call passes through one safety layer. The maker describes it as local, localhost-only by default, and without telemetry. It offers seven recommended compound tools and 98 granular tools for compatibility and debugging. Listed MCP hosts include Claude Code, Cursor, Windsurf, OpenClaw, and Zed; agents connect through stdio MCP or an HTTP MCP daemon. The maker says it works with any model. Clawd Cursor supports Windows x64 and ARM64, macOS 12 or later, and Linux X11 or Wayland. It is MIT licensed. Installation requires Node.js 20 or later, and desktop control requires one-time consent. On macOS, Accessibility permission is required; Screen Recording is optional for vision fallback.

Who it is for

It suits developers who want agents to handle desktop tasks locally when a native API or CLI is unavailable. It is available for Windows, macOS, and Linux users, subject to its installation and permission requirements.

What is good

Local operation is localhost-only by default, without telemetry.
Screen maps combine accessibility data and OCR.
Actions can be checked against expected outcomes.
Works with Windows, macOS, and Linux.
Free and MIT licensed.

What to know first

Requires Node.js 20 or later.
Desktop control requires one-time consent.
macOS requires Accessibility permission.

Verdict

Clawd Cursor focuses on local agent control of GUI tasks, with screen mapping and action checks. Check the Node.js and permission requirements before setting it up.

Clawd Cursor plans and pricing

All plans

Clawd Cursor Free Open source · MIT clawdcursor.com · 29 Sept 2026

Compared on AI computer use agents

Free plan: Yes
Desktop app control: Yes
Supported platforms: Windows, macOS, Linux
Browser control: Yes
Screen understanding: Yes
Deployment: local
Human approval controls: Yes
API access: Yes

Best Clawd Cursor alternatives

See all 20

Two bars Claude Code Free tier No. 1$17/mo6.8 Two bars Gemini Free tier No. 2Free6.6 Two bars QuickDesk Free tier No. 3Free6.6 One bar Lapu AI Free tier No. 5€16/mo6.5 One bar ChatGPT for Excel Free tier No. 6Free6.4 One bar Verdesk Free tier No. 7Free6.4

The record

What it does

Clawd Cursor gives AI agents local control of desktop apps through a screen map, actions such as clicks and typing, and action verification.

Screen reading

It combines accessibility data and OCR into a confidence-scored map with stable element IDs, using screenshots only when needed.

Safety

Every tool call passes through one safety layer, and actions can be checked against an expected outcome.

Local operation

The maker describes it as local, localhost-only by default, and without telemetry.

Integrations

The site lists Claude Code, Cursor, Windsurf, OpenClaw, and Zed as MCP hosts.

Tools

It offers 7 recommended compound tools and 98 granular tools for compatibility and debugging.

Operating systems

It supports Windows x64/ARM64, macOS 12 or later, and Linux X11/Wayland.

Connection modes

Agents can connect through stdio MCP or an HTTP MCP daemon on 127.0.0.1:3847/mcp.

Model support

The maker says it works with any model; the daemon can also use a configured built-in LLM.

Installation requirement

The GitHub README says Node.js 20 or later is required.

Permissions

Desktop control requires one-time consent; macOS also requires Accessibility permission, while Screen Recording is optional for vision fallback.

Intended use

The README describes it as a last-mile option for GUI tasks where a native API or CLI is unavailable.

License

The site identifies Clawd Cursor as MIT licensed.

Price from

Free