> ## Documentation Index
> Fetch the complete documentation index at: https://docs.claude-mem.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Screenpipe Integration

> Bring selected screen and meeting context into claude-mem with the bundled Screenpipe skill.

[Screenpipe](https://screenpipe.com) records screen text and audio on your computer.
The bundled `screenpipe` skill lets you recall that context from Claude Code and
save selected facts into claude-mem for future sessions.

## Setup

1. Install [Screenpipe](https://screenpipe.com), grant its recording permissions,
   and leave it running on the same computer as Claude Code.
2. [Install claude-mem](/installation) and restart Claude Code. The skill ships
   inside the plugin; no additional MCP server or package is required.
3. Invoke the skill:

```text theme={null}
/claude-mem:screenpipe Find the release discussion from my meetings this morning.
```

To save useful context, include that in your request:

```text theme={null}
/claude-mem:screenpipe Find yesterday's API design discussion and remember the decisions in this project.
```

The skill searches a bounded time window, summarizes matching screen text or audio
transcripts with source IDs and timestamps, and saves a selected note only when
asked. Saved notes use the existing local worker observation API and can be
retrieved with claude-mem's memory search tools.

```text theme={null}
Before: claude-mem recalls your agent sessions.
After:  Screenpipe screen/audio search -> selected context -> claude-mem observation
```

## Configuration and privacy

The helper requires Node.js 20+ (already required by claude-mem). It defaults to
Screenpipe at `http://127.0.0.1:3030`. Set `SCREENPIPE_URL` in the agent's environment
for another local port, and `SCREENPIPE_API_KEY` if Screenpipe requires bearer
authentication. Keep the API key out of prompts and command arguments.

Searches default to screen text and 20 results per page, with an explicit time
window. Audio is requested for meeting queries. The helper returns text and
source metadata, not screenshots or audio files. Long text is marked as truncated.

The HTTP requests go to local services, but **retrieved text enters the agent's
conversation** and is sent to its configured AI provider. Search results are
wrapped in `<private>` so claude-mem's automatic capture doesn't store them; only
a note you ask to save is stored. Saved notes are stored like any other
claude-mem memory: in the local database, and in your cloud account if cloud
sync is on. Choose a narrow topic, app, and time window.

Saving uses claude-mem's running local worker, found from `CLAUDE_MEM_WORKER_PORT`
or `~/.claude-mem/settings.json`. Hosted server mode and remote Screenpipe
instances are not supported by this basic integration. It does not enable
recording, periodically poll, or bulk-sync history.

## Troubleshooting

* **Connection failure or timeout:** check that Screenpipe or the claude-mem
  worker is running at the configured address. The helper exits with an error;
  it does not report an empty search or successful import.
* **HTTP 401/403:** check Screenpipe's API-key settings and `SCREENPIPE_API_KEY`.
* **No matches:** confirm Screenpipe recorded the requested interval and check
  the time zone, content type, and app filter.
* **Unconfirmed save:** keep the selected note and check memory before retrying,
  since a connection failure can happen after the worker stored it.

For a support report, include the operation and helper error; omit recorded text
and credentials unless you intend to share them.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.