screenpipe integrations
Give your AI tools the context behind the work
screenpipe is the local capture and search layer. Connect an AI tool through MCP, then retrieve the screen text, meeting, error, or decision that matters to the current task.
Current guides
Start with two integrations that solve different jobs
These are not product-name swaps. Each guide uses the current setup path, product-specific workflows, example prompts, and data boundaries for that tool.
screenpipe × Claude
Claude Desktop + Claude Code
Give Claude the work context you already have
Recall meetings, resume research, find an exact passage, or turn repeated work into a Pipe from Claude Desktop or Claude Code.
Open the setup guide
screenpipe × Codex
Codex app + CLI
Give Codex the context around the code
Reconnect requirements to implementation, recover transient errors, and resume work after a long or compacted coding session.
Open the setup guideCapture and index locally
screenpipe records screen and audio, extracts searchable text, and keeps the local index on the device in local-only mode.
Retrieve the relevant slice
The agent asks the local MCP server for a time range, phrase, meeting, app, or memory instead of receiving the full history.
Use it inside the current task
Claude or Codex receives the selected result in its context and can reason over it with the permissions you gave that tool.
The Connections page detects supported tools and keeps each setup separate, so you can connect only the agents you use.
More MCP guides
Using another coding tool?
The same local context layer also works with MCP-compatible editors. Existing guides cover Cursor and Windsurf.