Cursor - AI coding with screen context

Connect screenpipe to Cursor to give this AI-powered IDE access to your screen history and coding context.

Cursor is an AI-first code editor built on VS Code. with screenpipe integration, Cursor can reference what you’ve been working on across all your apps, not just the current editor.

one-click setup

the fastest way to connect screenpipe to Cursor:

1. Click the deeplink

   click here to install in Cursor

   this opens Cursor and automatically configures screenpipe as an MCP server.

2. Confirm installation

   Cursor will prompt you to confirm. click install or allow.

3. Test it

   in Cursor’s chat, try: “what was I working on this morning?”

ensure screenpipe is running before using Cursor with screenpipe features.

manual setup

if the deeplink doesn’t work, configure manually:

1. Open Cursor settings

   click Cursor → Settings → MCP (or search for “MCP” in settings)

2. Add new MCP server

   click add new global MCP server

3. Configure screenpipe

   • name: screenpipe
   • command: npx
   • args: ["-y", "screenpipe-mcp"]
4. Reload Cursor

   restart Cursor or reload the MCP servers

configuration file location

Cursor stores MCP config in .cursor/mcp.json in your project root, or globally:

{
  "mcpServers": {
    "screenpipe": {
      "command": "npx",
      "args": ["-y", "screenpipe-mcp"]
    }
  }
}

usage examples

once configured, Cursor can access your screen history naturally:

recall what you were reading:

> I was looking at documentation about async/await patterns earlier,
> find it and help me apply those patterns here

reference error messages:

> what error messages have I seen in my terminal today?

find API responses:

> find the API response format I was looking at in the browser
> and use it to type this function

meeting context:

> what did we discuss about authentication in the standup?
> help me implement those changes

clipboard history:

> I copied some code earlier but lost it, can you find it?

available tools

Cursor gets access to these screenpipe tools:

search-content

search your screen and audio history:

• filter by content type (all, ocr, audio, input, accessibility)
• time range filtering
• app/window filtering
• speaker filtering for audio
• full-text search

export-video

create video exports from screen recordings.

activity-summary

get app usage and activity stats.

list-meetings

list detected meetings with context.

see MCP server documentation for full details.

composer mode with context

Cursor’s Composer mode works great with screenpipe:

1. open Composer (Cmd+I or Ctrl+I)
2. reference screenpipe context in your prompt:
   > use screenpipe to find the authentication flow I was looking at,
   > then create a similar implementation here
   
3. Cursor will search screenpipe and use that context to write code

chat with context

use screenpipe in regular Cursor chat:

> @screenpipe what was I working on before lunch?

> search my screen history for the React component I was reading about

> find mentions of "database schema" in my meetings today

example workflows

context-aware coding:

user: I was reading about error handling patterns in Go, apply those here

Cursor: [searches screenpipe, finds content, applies patterns]

recall implementation details:

user: there was a TypeScript type definition I saw earlier that would work here

Cursor: [searches screen history, finds type, suggests implementation]

meeting follow-up:

user: implement the API changes we discussed in standup

Cursor: [searches audio transcriptions, extracts requirements, writes code]

tips for best results

be specific about what you want:

✅ "find the React component I was reading about this morning"
✅ "show me error messages from my terminal in the last hour"
✅ "what did we discuss about the API in today's meeting"

❌ "find that thing I saw" (too vague)

mention the app when relevant:

✅ "what was I looking at in Chrome"
✅ "find code from VS Code window"
✅ "show me Slack messages about the deploy"

use time ranges:

✅ "this morning"
✅ "in the last 2 hours"
✅ "yesterday afternoon"
✅ "during my 3pm meeting"

project-specific vs global

Cursor supports both project-level and global MCP servers:

project-level (.cursor/mcp.json in project root):

• only active in that specific project
• can customize per-project

global (via settings UI):

• active across all projects
• convenient for always-on tools like screenpipe

recommendation: use global for screenpipe so it’s available everywhere.

troubleshooting

MCP not showing up in Cursor?

• verify screenpipe is running: curl http://localhost:3030/health
• check MCP settings in Cursor to see if screenpipe is listed
• try restarting Cursor

Cursor can’t access screenpipe data?

• ensure screenpipe has screen recording permissions
• check screenpipe has been running and recording
• verify the time range you’re querying has data

queries timing out?

• reduce the time range
• be more specific with filters (app name, content type)
• avoid requesting screenshots unless needed

Node.js errors?

• ensure Node.js >= 18.0.0 is installed
• try npx clear-npx-cache and reinstall

deeplink not working?

• manually configure via settings instead
• check Cursor is up to date

still stuck? ask in our discord — we can help troubleshoot.

resources

• Cursor documentation
• MCP server documentation
• screenpipe API reference