Skip to main content
screenpipe acts as a memory layer for AI, giving large language models context about what you’ve been doing on your computer.

​
why AI needs memory

LLMs are powerful but stateless - they don’t know what you were working on 5 minutes ago. screenpipe bridges this gap by:
  • capturing everything on your screen
  • extracting text via OCR
  • transcribing audio
  • making it all searchable and retrievable

​
connect to AI providers

​
ollama (local)

# start ollama
ollama run phi4:14b-q4_K_M

# configure in screenpipe settings
# model: phi4:14b-q4_K_M
# url: http://localhost:11434

​
openai

set your API key in screenpipe settings under AI configuration.

​
claude via MCP

screenpipe can act as an MCP server, giving Claude direct access to your screen history: 
{
  "mcpServers": {
    "screenpipe": {
      "command": "screenpipe-mcp",
      "args": []
    }
  }
}
see MCP server setup for details.

​
build AI-powered apps

​
context-aware chatbot

import { pipe } from "@screenpipe/js";
import { generateText } from "ai";

async function chatWithContext(userMessage: string) {
  // get recent screen context
  const context = await pipe.queryScreenpipe({
    limit: 20,
    contentType: "all"
  });

  // format context for LLM
  const screenContext = context.data
    .map(item => {
      if (item.type === "OCR") {
        return `[Screen - ${item.content.app_name}]: ${item.content.text}`;
      } else {
        return `[Audio]: ${item.content.transcription}`;
      }
    })
    .join("\n");

  // generate response with context
  const response = await generateText({
    model: yourModel,
    prompt: `Based on the user's recent activity:\n${screenContext}\n\nUser question: ${userMessage}`
  });

  return response;
}

​
automatic note-taking

// obsidian pipe example - syncs screen activity to notes
import { pipe } from "@screenpipe/js";

async function logToObsidian() {
  const recentActivity = await pipe.queryScreenpipe({
    startTime: new Date(Date.now() - 60 * 60 * 1000).toISOString(), // last hour
    contentType: "all"
  });

  // summarize with AI and write to Obsidian
  // ... see obsidian pipe for full implementation
}

​
use cases

use casehow screenpipe helps
”what was I working on?“search screen history by time
”summarize today’s meetings”query audio transcriptions
”find that code snippet”search OCR text by content
”what did john say about X?“search transcriptions by keyword

​
privacy-first AI

  • all data stays on your device
  • use local LLMs (Ollama, LMStudio) for complete privacy
  • no data sent to cloud unless you explicitly choose cloud providers
  • filter what gets captured with window inclusion/exclusion

​
next steps