For founders, PMs & AI-native devs

Your meetings as long-term context for Claude, Cursor, and Codex.

Your AI already knows your codebase. Quietly adds your standups, 1:1s, and customer calls as plain Markdown in a folder your tools already index. A local MCP server ships next.

Download for Mac See the stack
Free 7-day trial Plain Markdown, your folder MCP shipping next $49 once
One block of config
Claude reads your meetings.

Add Quietly to claude_desktop_config.json. Claude calls search_meetings and get_transcript as tools.

{ "mcpServers": { "quietly": { "command": "quietly-mcp", "args": ["--vault", "~/Notes/Meetings"] } } }
The bottleneck

Your AI knows your repo. It still doesn’t know your meetings.

You paste transcripts into Claude. The context vanishes when the chat ends, with no reuse, no recall, and no thread back to the source.

Granola encrypted its local DB. The community had to reverse-engineer it to get the notes back out. Then the format changed.

Otter and Fireflies meter the AI you already pay for. Your transcript, their chatbot tier, their prompts.

Your standups should be a tool call. A source the agent reaches for on its own, instead of a screenshot or a paste.

The fix

One folder, one MCP server, and the AI tools you already use.

MCP shipping next

Claude Desktop talks to your meetings.

Today its filesystem MCP reads your notes folder. Ask “what did we decide about the Q3 launch?” and it finds the right file. A first-class Quietly MCP server ships next.

Cursor & Codex CLI

@meeting in your IDE or terminal.

Reference any past meeting from inside Cursor while you code, or grep it from the terminal with Codex CLI. Pull a planning call into the chat as you implement against it.

Plain Markdown

Grep it, script it, keep it.

Notes save to a folder you pick: iCloud, the repo your code lives in, anywhere on disk. Front-matter has attendees, calendar, and duration. The body is the clean, LLM-polished transcript.

On-device

Standups stay between you.

Whisper large-v3 and LLM cleanup run on your Mac, so no audio uploads. Quietly deletes the audio after transcription unless you keep it.

Calendar-aware

Record without remembering to.

Quietly reads Google or Apple Calendar and prompts you before recordable meetings. Press ⌘R for ad-hoc capture, or pick Transcribe a file… for an existing audio file. Same Markdown, same folder.

$49 once

The one tool in your stack that doesn’t bill monthly.

One-time license. Keep the version you bought. No AI credits, no per-meeting fee, no procurement review next quarter.

Your stack

Where it slots into the tools you already pay for.

Quietly is the data layer. Your AI tools are the interface.

Claude Desktop
Reads search_meetings, get_transcript, ask_meeting as tools
MCP
Cursor
Reference past meetings as @meeting while you code
MCP
Codex CLI
Same MCP tools, called straight from your terminal session
MCP
iCloud · Git · Dropbox
Plain .md on disk. Sync it however you already do
folder
VS Code · ripgrep · your scripts
Plain .md. Pipe it into anything that reads files.
fs
Anthropic · OpenAI · Ollama
BYOK summarisation. You choose the prompt and the model.
BYOK
Why now

MCP turns your meeting notes into a tool your AI can call.

With MCP, your meetings become a first-class tool inside the AI you already use, as long as the data is yours to expose. Markdown in your folder is the most boring, durable, agent-friendly format there is. Quietly writes it.

FAQ

Things builders ask before installing.

How will the MCP server work?
Quietly is shipping a local MCP server (quietly-mcp) that exposes three tools to any MCP-aware client such as Claude Desktop, Cursor, or Codex CLI: search_meetings, get_transcript, and ask_meeting. The AI calls them as tool calls, the same way it calls the filesystem or Git MCP servers. You add one block to your client config, and the client finds your meetings on its next launch. MCP support ships next. Today, notes are plain Markdown that your editor's filesystem MCP or file readers already pick up.
Does it leak my notes to Anthropic / OpenAI?
Only when you ask Claude, Cursor, or Codex a question that pulls in a meeting. The transcript then flows through whichever provider that AI uses, the same as if you had pasted it in by hand. Quietly itself never sends notes anywhere.
What about summaries?
Quietly's built-in Chat writes them with your own key. Connect Anthropic or OpenAI (the key stays in your Keychain) and ask for a summary, action items, or decisions. Each answer cites the exact transcript segment. On-device cleanup needs no key. If you'd rather drive it yourself, every note is plain Markdown, so you can point Claude, ChatGPT, Cursor, or Codex at it any time.
Where do the files live? Will my sync break?
You pick the folder: your repo, iCloud, Dropbox, anywhere on disk. The notes are plain .md, so Git, ripgrep, and your own scripts all work, and iCloud and Dropbox sync them like any other file.
Can I script post-processing, like pushing to Linear?
Yes, and that is the design. Quietly writes a Markdown file to disk and emits a notification. From there it's your shell: an fswatch hook, a Hammerspoon script, an n8n workflow, a Cursor fix-up pass on the transcript. We don't ship integrations because Markdown plus your own scripts is more flexible.
Do you have a team license for a small founding team?
Team licensing isn't a public SKU yet. We're sizing it with early teams, so email support@meetquietly.com with your headcount and we'll sort out multi-seat licensing on a single invoice. For now each person buys a Personal license ($49, 2 Macs) and points Quietly at their own folder. There is no shared cloud, so coordination happens through whatever you already use: Linear, Notion, or a shared notes folder in Git.

Wire your meetings into your AI.

Free 7-day trial · No account, no card. Then $49 once, a line item that stays the same next quarter. macOS 13+.

$ 49 once · 2 Macs
Download for Mac

Buying for a team? Contact us