Overheard
FreeNot checkedMac menu bar meeting recorder: records calls without bots, transcribes (cloud or on-device), and makes your meetings searchable from Claude via MCP.
About
Mac menu bar meeting recorder: records calls without bots, transcribes (cloud or on-device), and makes your meetings searchable from Claude via MCP.
README
Overheard
Your meetings, turned into text — and answerable from Claude.
A tiny Mac menu bar app that records your calls, transcribes them and makes them searchable from Claude.
No bot joins your call. No virtual audio driver. Native Swift.

getoverheard.dev · ▶ Watch with sound · Download
Overheard is a free, open-source meeting recorder for Mac. It lives in the menu bar, records Google Meet, Microsoft Teams and Slack calls without a bot joining the meeting, transcribes them with speakers and timestamps (on your Mac or in the cloud), and lets Claude Code, Claude Desktop or Codex answer questions about your meetings through an MCP server. A free, local alternative to Granola, Otter and Fireflies.
How it works
![]() |
![]() |
| 1 · Record. It starts when Meet, Teams or Slack grab the mic. Bookmark moments with ⌃⌥M. | 2 · Transcribe. Every word, with speakers and timestamps — in the cloud or fully on your Mac. |
![]() |
![]() |
| 3 · Summarize. Decisions and action items you can tick off. | 4 · Ask Claude. "What did we decide about the Acme launch?" — answered from your meetings, with timestamps. |
Why Overheard
- Your meetings, inside Claude. Overheard ships an MCP server: ask Claude Code, Claude Desktop or Codex "what did we decide in yesterday's standup?" or "draft the follow-up email for the Acme call" and it reads your transcripts.
- No extra API keys for the AI. Summaries and chat use the Claude Code or Codex CLI you already have, with your existing subscription.
- Records the call, not the meeting. System audio is captured with Core Audio process taps (macOS 14.2+), your voice from the microphone on a separate track. Works with Google Meet, Microsoft Teams, Slack huddles, and anything else that plays audio.
- Cloud or fully local transcription. ElevenLabs Scribe for the best accuracy, or Cohere Transcribe running on your Mac's GPU if audio must never leave the machine.
Features
Recording
- Starts automatically when Meet, Teams or Slack start using the microphone (or with one click, or
overheard://toggle) - Bookmark important moments while recording with ⌃⌥M from any app
- Your microphone is leveled to match the call, with optional echo reduction for people who use speakers
- Calendar integration: meetings get the event title and attendees
- Survives device changes (AirPods connecting mid-call, sample rate switches)
Transcripts
- Speaker diarization; your own voice is labelled with your name
- Rename speakers, edit lines, click any line to play it back
- Talk-time per speaker, search with highlighting, export audio sections as
.m4a - Import audio/video files or links (YouTube, podcasts — via
yt-dlp) and transcribe them too
AI
- Automatic summary with decisions, action items (checkable) and bookmarked moments
- Chat across all your meetings or a single one
- MCP server with
list_meetings,search_meetings,get_meeting,list_action_itemsand, if you allow it, actions like starting a recording or checking off a task
Sharing
- PDF export, copy summary/transcript, markdown notes with frontmatter for Obsidian, Logseq or Notion
The interface is in English and Italian.
Install
Requires an Apple Silicon Mac with macOS 14.2 or later (on-device transcription needs macOS 15).
- Download the latest
.dmgfrom Releases and drag Overheard to Applications — orbrew install --cask anonbeat1/tap/overheard. It is signed and notarized by Apple, so it opens with a double click. - Open it: the icon appears in the menu bar. Grant microphone and audio recording access on the first recording.
- Transcription runs On this Mac by default on macOS 15 (downloads ~2.5 GB of models once, no account). For the best accuracy, switch to ElevenLabs in Settings → Transcription and paste an API key with the Speech to Text permission.
- Optional: install Claude Code or Codex for summaries and chat, then click Connect to Claude in Settings to enable the MCP server.
Claude Code plugin
Instead of Connect to Claude, Claude Code users can install the plugin: the same MCP server plus three skills.
/plugin install overheard --marketplace anonbeat1/overheard
/overheard:recap— decisions, action items and open questions of a meeting, with the minute of each/overheard:followup— a ready-to-send follow-up email/overheard:todo— your open action items across meetings
It expects the app in /Applications. Use one or the other: with both, Claude sees the tools twice.
Build from source
git clone https://github.com/anonbeat1/overheard.git
cd overheard
./build.sh # builds the app and the local transcriber, installs to ~/Applications
./package.sh # optional: creates dist/Overheard-<version>.dmg
Needs Xcode 16 or later. The first build downloads and compiles the on-device transcription stack (MLX), which takes a few minutes.
Where your data lives
Everything stays in ~/Meetings, one folder per meeting:
~/Meetings/2026-10-14_113035/
recording.wav call audio
mic.wav your microphone
transcript.md **[mm:ss] Speaker:** text
summary.md summary, decisions, action items
bookmarks.txt bookmarked moments (seconds)
Plain files, so other tools can read them. Audio is sent to ElevenLabs only if you choose it as the engine; summaries and chat send transcript text to the AI CLI you use. Old meetings can be moved to the Trash automatically after 7–365 days.
Automation
| URL | Action |
|---|---|
overheard://toggle |
start / stop recording |
overheard://bookmark |
bookmark the current moment |
overheard://meetings |
open the meeting library |
overheard://import?url=<link or file> |
import and transcribe |
open -g overheard://toggle
FAQ
Is Overheard free? Yes. MIT licensed, no account, no subscription. ElevenLabs bills you for cloud transcription only if you choose it; on-device transcription costs nothing.
Does a bot join the meeting? No. Overheard records the call from your Mac with Core Audio process taps, so nothing appears in the participant list.
Does it work with Zoom or Webex? Yes, it records anything that plays audio on your Mac. It starts automatically for Meet, Teams and Slack; for other apps, start it with one click.
Does it run on Intel Macs or Windows? No. It needs an Apple Silicon Mac with macOS 14.2 or later (macOS 15 for on-device transcription).
Do I need Claude? No. Recording and transcription work on their own; summaries and chat use Claude Code, Codex or Apple Intelligence, and the MCP server is for asking from Claude.
Please record responsibly
Recording people may require their consent, depending on where you and they are. Tell participants you are recording. You are responsible for complying with the laws that apply to you.
Credits
- speech-swift (Apache 2.0) for on-device inference
- Cohere Transcribe (Apache 2.0), MLX conversion by aufklarer
- pyannote speaker-diarization-community-1 (CC-BY-4.0) and segmentation-3.0 (MIT) by pyannote, Core ML / MLX conversions by aufklarer
- ElevenLabs Scribe for cloud transcription
Not affiliated with Anthropic, OpenAI, Google, Microsoft, Slack or ElevenLabs.
License
MIT © 2026 Gabriele Spilotri Cavaiola
Installing Overheard
This server has no published package — it is built from source. Open the repository and follow its README.
▸ github.com/anonbeat1/overheardFAQ
Is Overheard MCP free?
Yes, Overheard MCP is free — one-click install via Unyly at no cost.
Does Overheard need an API key?
No, Overheard runs without API keys or environment variables.
Is Overheard hosted or self-hosted?
Self-hosted: the server runs locally on your machine via the install command above.
How do I install Overheard in Claude Desktop, Claude Code or Cursor?
Open Overheard on unyly.org, pick your client tab (Claude Desktop, Claude Code, Cursor) and press Install — the config is generated automatically, no JSON editing.
Related MCPs
ARA
Generate images, video and audio from any AI agent — one connector.
by ARAOmni Video
An MCP server that transforms LLM-enabled IDEs into professional video editors by pre-processing footage into text proxies, generating motion graphics via HTML/
by buildwithtazaYouTube
Transcripts, channel stats, search
by YouTubeEverArt
AI image generation using various models.
by modelcontextprotocolgpu-bridge/mcp-server
GPU inference API for LLMs, image and video generation, TTS, Whisper, embeddings, reranking and OCR, paid per use via x402 or an API key.
by gpu-bridgemerterbak/Grok-MCP
Access xAI's Grok API with agentic tool calling, image generation, vision and file support.
by merterbakSureScaleAI/openai-gpt-image-mcp
Image generation and editing with OpenAI's GPT image models.
by SureScaleAIYangLiangwei/PersonalizationMCP
Aggregate personal data from Steam, YouTube, Bilibili, Spotify, Reddit and other gaming, music, video and social platforms, with OAuth2 and token management.
by YangLiangweiAceDataCloud/MCPFlux
Flux image generation and editing from Black Forest Labs via the Ace Data Cloud API.
by AceDataCloudAceDataCloud/MCPNanoBanana
Nano Banana image generation and editing, including virtual try-on and product placement in realistic scenes.
by AceDataCloudCompare Overheard with
Not sure what to pick?
Find your stack in 60 seconds
Author?
Embed badge for your README
Browse similar
All media MCPs




