Product ·
Connect Your Calendars to Oatmeal
How to get Google, Outlook, and iCloud calendars into Oatmeal in under two minutes.
Free for macOS
Meeting notes that
never leave your Mac.
On-device transcription and AI. Search across every conversation you've ever had. Free. Runs entirely on your Mac.
macOS 15 · Apple Silicon · Free Forever
Capabilities
Perfect recall.
Privacy
Entirely on your Mac
Transcription, summarization, and search all run locally. Your audio never touches a server. Not even once.
Intelligence
Memory across meetings
Search understands entities, context, and time—not just keywords.
Access
Free forever
No subscription, no account, no usage limits. Just download and use it.
Invisible
No bot in the room
Captures audio directly from macOS. No awkward consent prompts.
Query
Cross-meeting queries
Press ⌘K and ask anything. Answers synthesized from every meeting, with citations.
Ingest
Import anything
Audio files, video, VTT/SRT subtitles, YouTube transcripts. Everything becomes searchable memory — processed locally, of course.
The Engine
How it works
1
Record
Press ⌘⇧M to capture any call. System audio and mic recorded separately.
2
AI processes locally
Parakeet transcribes. Qwen3 summarizes. Everything on Apple Silicon. Nothing uploaded.
3
Search everything
Hybrid search finds meetings by meaning. Ask questions across months with ⌘K.
Nothing leaves your Mac.
Every privacy claim is backed by architecture. No cloud calls, no telemetry, no data collection.
100% local architecture
No network calls, no cloud APIs, no server-side processing.
No telemetry
Zero analytics, zero tracking. We don't know who uses it.
No account required
Download and run. No signup, no email, no password.
No internet required
Works perfectly offline. On a plane, in a basement, anywhere.
Audio stored locally
Raw audio files live in your Documents folder. Your data, your device, your control.
FAQ
Frequently asked questions
Is it really completely local?
Yes. Oatmeal uses Parakeet TDT 0.6B-v3 for transcription and Qwen3-4B for AI features. Both run on your Mac via Apple's MLX framework and CoreML. No API calls, no cloud processing.
What Macs are supported?
Any Mac with Apple Silicon (M1 or later) running macOS 15+. 8GB RAM minimum, 16GB+ recommended. Intel Macs are not supported.
How does it compare to Granola?
Similar features: transcription, AI summaries, cross-meeting search. Key differences: Oatmeal runs locally (Granola uses cloud AI), Oatmeal is free (Granola is $14–35/mo after 25 meetings), and Oatmeal includes system-wide dictation. Full comparison.
Will it keep working if you disappear?
Yes. Oatmeal runs completely offline with no dependency on our servers. Even if we vanish tomorrow, your app keeps working. That's the point.
What platforms does it record from?
Oatmeal captures system audio directly from macOS — it works with Zoom, Google Meet, Teams, Webex, FaceTime, phone calls, or any app that plays audio through your Mac. No meeting bot required.
How much storage does it need?
The app is about 3GB including the AI models. Audio recordings use roughly 60–100MB per hour and are stored in your Documents folder. Transcripts and metadata are lightweight — a year of daily meetings typically uses under 1GB.
Can I export my meeting notes?
Copy-paste works for any transcript or summary. Structured export to Markdown and PDF is on the roadmap.
Is my data backed up?
Your data lives in standard files on your Mac — audio in Documents, database in Application Support. Back them up with Time Machine, iCloud Drive, or any backup tool. Since nothing is in the cloud, your backup strategy is your own.
Blog
From the Blog
Deep Dive ·
Search That Actually Understands Your Meetings
How Bayesian probability fusion replaces ad-hoc ranking to combine keywords, semantics, entity awareness, and temporal context into calibrated meeting search.
Deep Dive ·
The Rise of Local Speech Recognition
How speech recognition escaped the cloud: from Hinton's neural networks to Whisper to 50x-faster-than-realtime local models. A 20-year technical history.
Engineering ·
Building Local-First in 2026
The gap between 'theoretically possible' and 'ships as a product' is enormous. Here's what we learned building on-device AI for meeting transcription.
Product ·
The Meeting Problem
Workers spend 57% of their day on communication. The issue isn't meetings themselves—it's the memory gap that makes us have the same conversations twice.
Product ·
The Bot in the Room
When a transcription bot joins your call, the room changes. The observer effect is real—and it's silently degrading conversations everywhere.