- Home
- AI Computer Use Agents
- Clawd Cursor
Clawd Cursor
What it is
Clawd Cursor gives AI agents local control of desktop apps for GUI tasks where a native API or command-line option is unavailable. It builds a screen map from accessibility information and OCR, with stable element IDs and confidence scores; screenshots are used only when needed. Agents can click and type, and actions can be checked against an expected outcome. The maker says each tool call passes through one safety layer and describes the software as local, localhost-only by default, and without telemetry. Connections use stdio MCP or an HTTP MCP daemon, and listed hosts include Claude Code, Cursor, Windsurf, OpenClaw, and Zed. Clawd Cursor includes seven recommended compound tools and 98 granular tools for compatibility and debugging; its maker says it works with any model. It supports Windows x64/ARM64, macOS 12 or later, and Linux X11/Wayland. The tool is free and MIT licensed. Installation requires Node.js 20 or later. Desktop control requires one-time consent; macOS also requires Accessibility permission, while Screen Recording is optional for vision fallback.
For and against
For
- Maps screen elements using accessibility data and OCR.
- Actions can be checked against expected outcomes.
- Supports Windows, macOS, and Linux.
- Free and MIT licensed.
Against
- Requires Node.js 20 or later.
- Desktop control requires one-time consent.
- macOS requires Accessibility permission.
The verdict
Clawd Cursor provides local desktop control with screen mapping and action checks for GUI-oriented agent tasks. Review the Node.js and permission requirements before setting it up.
Who it suits
Clawd Cursor suits people building AI-agent workflows that need desktop GUI interaction where a native API or CLI is unavailable. It may fit users working with the listed operating systems and MCP hosts who can meet its installation and permission requirements.
Plans and prices
Clawd Cursor
Open source · MIT
Read 29 Sep 2026 · clawdcursor.com
On the record
- Connection modes
- Agents can connect through stdio MCP or an HTTP MCP daemon on 127.0.0.1:3847/mcp.clawdcursor.com · read 29 Sep 2026
- Installation requirement
- The GitHub README says Node.js 20 or later is required.github.com · read 29 Sep 2026
- Integrations
- The site lists Claude Code, Cursor, Windsurf, OpenClaw, and Zed as MCP hosts.clawdcursor.com · read 29 Sep 2026
- Intended use
- The README describes it as a last-mile option for GUI tasks where a native API or CLI is unavailable.github.com · read 29 Sep 2026
- License
- The site identifies Clawd Cursor as MIT licensed.clawdcursor.com · read 29 Sep 2026
- Local operation
- The maker describes it as local, localhost-only by default, and without telemetry.clawdcursor.com · read 29 Sep 2026
- Model support
- The maker says it works with any model; the daemon can also use a configured built-in LLM.clawdcursor.com · read 29 Sep 2026
- Operating systems
- It supports Windows x64/ARM64, macOS 12 or later, and Linux X11/Wayland.clawdcursor.com · read 29 Sep 2026
- Permissions
- Desktop control requires one-time consent; macOS also requires Accessibility permission, while Screen Recording is optional for vision fallback.github.com · read 29 Sep 2026
- Safety
- Every tool call passes through one safety layer, and actions can be checked against an expected outcome.clawdcursor.com · read 29 Sep 2026
- Screen reading
- It combines accessibility data and OCR into a confidence-scored map with stable element IDs, using screenshots only when needed.clawdcursor.com · read 29 Sep 2026
- Tools
- It offers 7 recommended compound tools and 98 granular tools for compatibility and debugging.clawdcursor.com · read 29 Sep 2026
- 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.clawdcursor.com · read 29 Sep 2026
