Pocket-Ingest pipes voice memos to Claude Code via iCloud
Pocket-Ingest is an open-source tool that uses an iOS Shortcut to record, transcribe, and save voice memos as `.txt` files to iCloud, which Claude Code then picks up automatically at the start of a session — no servers, apps, or API keys required.
Score breakdown
Capture async voice context on mobile and feed it directly into Claude Code sessions without any infrastructure overhead.
- 01An iOS Shortcut records a voice memo, transcribes it, and saves a `.txt` file to iCloud.
- 02Claude Code automatically picks up the file at the start of the next session.
- 03No servers, apps, or API keys are required; setup is described as five minutes.
Pocket-Ingest, shared by filipeisho on Hacker News, addresses a gap in mobile-to-desktop AI workflows. The author describes using what they call "Karpathy's LLM wiki pattern" — raw sources stored in a vault with Claude maintaining a knowledge base on top — which works well at a desk but breaks down when away from it. Pocket-Ingest fills that gap by letting an iOS Shortcut record a voice memo, transcribe it, and write a `.txt` file to iCloud, which Claude Code then automatically ingests at the start of the next session.
The tool is intentionally minimal: no servers, no dedicated apps, and no API keys are required, with a claimed five-minute setup time.
The tool is intentionally minimal: no servers, no dedicated apps, and no API keys are required, with a claimed five-minute setup time. The current implementation is strictly one-directional — from phone to computer. The author flags two-way communication as a more interesting future direction but chose to ship the simplest version first to gauge community interest.
Key facts
- 01An iOS Shortcut records a voice memo, transcribes it, and saves a `.txt` file to iCloud.
- 02Claude Code automatically picks up the file at the start of the next session.
- 03No servers, apps, or API keys are required; setup is described as five minutes.
- 04The current flow is one-way: phone to computer only.
- 05The author describes the broader workflow as 'Karpathy's LLM wiki pattern', using a vault of raw sources with Claude maintaining a knowledge base on top.
- 06The author expressed interest in eventually building two-way communication between phone and Claude Code.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content.