GitHubMIT

MCP server for desktop AI

Give AI agents access to your screen history and audio transcriptions. screenpipe implements the Model Context Protocol for seamless AI integration.

📧
🌐
💬
📹
📝
🖥️
capturing everything

Quick answer

screenpipe MCP server gives AI agents access to your screen history and desktop context. Works with Claude Desktop and MCP-compatible tools.

See it in action

AI agents are context-starved

Agents can't help with what they can't see.

01

Claude Desktop can't access your screen history

02

AI agents have no memory between sessions

03

Context gets lost across conversations

04

Building custom integrations is complex

05

No standard way to give AI desktop context

MCP: The standard for AI tools

screenpipe implements Model Context Protocol (MCP), Anthropic's standard for connecting AI to tools. Install once, every MCP-compatible agent gets access to your screen history.

Screen history search

AI agents can search your screen text and audio transcriptions. Find anything you've seen or heard.

Meeting context

All your meetings transcribed and searchable. Ask about past discussions.

Full workflow memory

Code, research, emails - everything captured. AI understands your complete context.

Video export

Export screen recordings as MP4 clips for sharing or analysis.

How it works

1

Install screenpipe

Download and run screenpipe to start building your screen history.

2

Download MCP server

In Settings → Connections, download the MCP server bundle.

3

Connect your AI

Claude Desktop and other MCP clients automatically discover screenpipe search tools.

Code examples

Available MCP tools

Tools exposed by the screenpipe MCP server

// Search screen history
{
  name: "search",
  description: "Search screen text and audio transcriptions",
  parameters: { query: string, contentType?: string, limit?: number }
}

// Search by time range
{
  name: "search_time_range",
  description: "Find content from a specific time period",
  parameters: { start: string, end: string, query?: string }
}

// Get recent context
{
  name: "recent_context",
  description: "Get what was on screen recently",
  parameters: { minutes?: number }
}

// Export recordings
{
  name: "export_video",
  description: "Export screen recording as MP4",
  parameters: { startTime: string, endTime: string }
}

Key benefits

Standard protocol - works with any MCP client
Rich desktop context for AI agents
Persistent memory across sessions
Open source and extensible
All data stays local

Frequently asked questions

Model Context Protocol (MCP) is an open standard from Anthropic for connecting AI assistants to external tools and data sources. It's like a USB for AI - plug in tools and any MCP-compatible AI can use them.

Claude Desktop is the primary MCP client. Other tools are adopting the standard, including Cursor and various AI agent frameworks. Any MCP-compatible client can use screenpipe tools.

screenpipe stores everything locally. When you search via MCP, only the search results you request are shared with the AI. Your full screen history stays on your machine.

Yes. screenpipe is open source and the MCP server code is available. You can extend it with custom tools or use the screenpipe API to build entirely new MCP servers.

Enable AI desktop context

The MCP server your AI agents need.