PilotGentic
Give your agent real macOS desktop and browser control—clicks, typing, screenshots, accessibility tree—for flows you cannot complete with API-only tools.
Overview
PilotGentic MCP is a MCP server for the Build phase that provides macOS desktop and browser automation tools (click, type, screenshot, AX tree) to coding agents over stdio.
What is this MCP server?
- macOS-only: click, type, screenshot, accessibility (AX) tree, and browser control
- Supports agentic loops for multi-step desktop automation from MCP
- npm package @pilotgentic/mcp-server (v1.0.0) with stdio transport
- Documented at pilotgentic.com with source under pilotgentic repo subfolder components/mcp-server
- Server version 1.0.0
- npm identifier: @pilotgentic/mcp-server
- Platform scope: macOS desktop control (per published description)
What problem does it solve?
API-only agents cannot operate native Mac apps or opaque web UIs, blocking solo builders from automating the last mile of manual QA and demos.
Who is it for?
Mac-based indie developers building agent products that need real UI interaction during development.
Skip if: Windows/Linux-only shops, unattended production bots without human oversight, or teams that only need read-only HTTP APIs.
What do I get? / Deliverables
After installing @pilotgentic/mcp-server, your agent can drive the Mac desktop and browser through MCP tools for hands-on build and validation loops.
- MCP tools for click, type, screenshot, AX inspection, and browser control on Mac
- Repeatable agentic desktop loops for build-time UI tasks
Recommended MCP Servers
Journey fit
Agent-tooling on Build is where you extend what the coding agent can physically do on a dev machine while implementing or validating UI-heavy products. PilotGentic is local control plumbing (AX tree, browser, loops), not distribution SEO or production error monitoring.
How it compares
Local macOS desktop MCP automation, not a cloud browser farm or CRM integration server.
Common Questions / FAQ
Who is com.pilotgentic/mcp-server for?
Solo builders on macOS using Claude Code or Cursor who need the agent to control the desktop, accessibility tree, and browser—not just edit repo files.
When should I use com.pilotgentic/mcp-server?
During Build and early Ship manual checks when automating repetitive UI steps, capturing screenshots, or running agentic loops on local apps.
How do I add com.pilotgentic/mcp-server to my agent?
Install @pilotgentic/mcp-server from npm, register the stdio server as com.pilotgentic/mcp-server in your MCP client, and grant macOS accessibility permissions PilotGentic requires.