AI agents are good at processing text, but they have no memory of what you actually said in meetings, calls, or hallway conversations. Hemory fills that gap: it listens continuously through your phone or Apple Watch, transcribes everything into searchable text, and connects the result to your AI agents via MCP.
| Category | AI Memory & Productivity |
|---|---|
| Website | hemory.com |
| Platforms | iOS, Android (macOS, Windows, Linux, Web coming soon) |
| Languages | 100+ languages supported |
| Agent Support | Claude Code, Codex, Gemini, Cursor, VS Code via MCP |
How It Works
Hemory runs in two modes. Manual mode records only when you tap start, giving you control over exactly what gets captured. Schedule mode runs automatically during configured hours (for example, weekdays 9:00 to 18:00), which is what the "always-on" in the tagline actually means.
Audio is processed as a stream, transcribed with speaker labels, and segmented into "moments" by activity. The raw audio is stored only on the local device and never syncs to the cloud. Once transcription is complete, the audio can be deleted automatically, so continuous listening doesn't fill your phone's storage.
The MCP Connection
The technical differentiator is MCP (Model Context Protocol) integration. After connecting Hemory to an AI agent, the agent can call search_memory to look through your transcribed conversations. Setup is two commands:
codex mcp add hemoryfor Codexclaude mcp add hemoryfor Claude Code
Once connected, the agent has access to everything Hemory has heard. You can ask it to draft a follow-up email from a meeting, compile a monthly work report from all your conversations, or find what someone promised you in last Tuesday's sync. The agent works from real conversational context rather than whatever you remembered to type into notes.
Practical Use Cases
The landing page shows three compelling examples: generating a monthly report deck from a month of work conversations, writing a nightly journal entry from the day's interesting moments, and drafting a PRD from customer interviews.
For founders and consultants who spend most of their day in meetings, the value proposition is clear: stop taking notes manually and let the agent reconstruct whatever you need from the source material.
Privacy Model
Hemory takes a privacy-first approach. Audio never leaves the device. Transcriptions are encrypted at rest and in transit. Processing happens as a stream with zero audio retention in the cloud. The on-device storage model means there's no server-side audio database to breach.
That said, MCP connections do send transcription data to whichever AI agent you connect. Users should understand that "private" means audio stays local, not that the text stays local once you pipe it to Claude or Codex.
Limitations
Mobile-first for now. Desktop and web apps are listed as "coming soon." If your important conversations happen at a desktop workstation, you're relying on your phone sitting nearby.
Battery and storage. Continuous listening on a phone has real hardware costs. While Hemory optimizes by deleting processed audio, the battery impact of always-on microphone access will vary by device.
Consent. Recording conversations raises legal and ethical questions in many jurisdictions. Hemory provides the tool, but the responsibility for informing participants falls on the user.
Pros
- Always-on listening with automatic segmentation
- MCP integration with major AI agents
- 100+ languages with speaker labeling
- Zero cloud audio retention
- Schedule mode for hands-free operation
Cons
- No desktop app yet (coming soon)
- Battery impact of continuous listening
- Recording consent is the user's responsibility
- Transcription data does leave device via MCP
Verdict
Hemory occupies a genuinely new category: it's not a note-taking app, not a meeting recorder, and not a transcription service. It's a memory layer that feeds AI agents with real-world conversational data. The MCP integration is the key technical decision, because it means any MCP-compatible agent immediately gains access to your conversation history without custom integration work. For professionals whose work is mostly talking, this is the bridge between what they said and what their AI can act on. The privacy model is thoughtful, though users should be clear-eyed about what "private" means once data flows to an external AI provider.
Try it at hemory.com.
Build AI tools?
Get a permanent feature on AI Tools Hub — indexed by Google, linked from reviews and roundups.
Get Featured