Listenr vs Anarlog
Pick Anarlog if anyone you work with is on Windows or Linux, if you want open source, or if you need webhooks and a hosted API; it is free and does a lot. Pick Listenr if you are on a Mac and want a native app that transcribes on the device without changing a setting, names speakers on Zoom, Google Meet and Microsoft Teams without manual labelling, and lets an assistant drive every command. Both record without a bot and ship a command line tool and a local MCP server for free.
| Feature | Listenr | Anarlog |
|---|---|---|
| Does a bot join the call | No, capture is from the Mac | No, capture is from the desktop |
| Where your audio is transcribed by default | On the Mac; four of five backends are offline | Either, and the setup suggests the hosted path first |
| Who writes the notes | A local model, Apple Intelligence, your own Anthropic or OpenAI key, or Listenr Cloud | The same range, plus an existing Claude or ChatGPT subscription with no API key |
| Speaker names | Automatic on Zoom (opt-in), Google Meet (extension) and Teams; Speaker 1, 2 elsewhere | Manual labelling free, better identification on Pro |
| What else it reads from the call | On Zoom and Google Meet: who joined and left, and the chat, fed into the notes | Not described |
| Devices you can use it on | macOS 15 or later, Apple Silicon | macOS; Windows, Linux, iOS and Android in beta |
| What the desktop app is | A native macOS app, written in Swift for the Mac alone | A Tauri app, a Rust core and a web view, one build shared across three platforms |
| Licence and price | Closed source. Pro $19/month, or $15/month billed yearly | MIT licensed and forkable. Pro $14/month or $140/year |
| What the free plan gives you | Unlimited recording and on-device transcription, speaker separation, local or own-key notes, CLI and MCP; personal use only | Unlimited on-device transcription, manual speaker labelling, your own keys, CLI, MCP and webhooks; no licence restriction |
| Is your content used for training | Local models send nothing; Listenr Cloud stores no transcript | Never used to train models, per its privacy policy |
| Getting at meetings from a terminal or an assistant | Every command as a CLI verb; a local MCP server, read-only by default, over 150 tools with control enabled | A CLI, 11 MCP tools with a proposal write path, webhooks and a hosted API |
| Where the recording ends up | Deleted after transcription unless you keep it; kept files stay on the Mac | On the device by default; cloud transcription routes the audio through its servers |
These two agree on more than any other pair on this site. Both record with no bot in the call, both can transcribe on the device, both let you choose the model that writes the notes, and both ship a command line tool and a local MCP server for free.
Listenr transcribes on the Mac unless you change it, names speakers on Zoom, Google Meet and Microsoft Teams without labelling, puts who joined and the meeting chat into the notes, and lets an assistant run every command. It is a native Mac app and nothing else.
Anarlog is free and open source under the MIT licence, runs on Windows and Linux, has phone apps in beta, and has webhooks and a hosted API that Listenr lacks. Listenr’s free plan is licensed for personal use only. If none of Listenr’s differences is worth a subscription to you, Anarlog is the answer.
The app was called Hyprnote, then Char, and took the name Anarlog in May 2026. The core team now builds char; the Anarlog community app remains open source, MIT licensed and maintained.
Where does the transcription run by default?
Both can transcribe on the device and both can send audio to a hosted service. What differs is where a new user lands.
Listenr’s default backend is Apple’s on-device engine, in one language you set. Parakeet, a download, adds 25 European languages with detection, and SenseVoice covers Chinese, Japanese and Korean. The one cloud backend, OpenAI, sends nothing until you pick it.
Anarlog’s on-device options are Apple Speech on macOS 26, Parakeet models for the same 25 European languages (live, or after the recording with speaker labels), a bundled model, or a whisper.cpp file you supply, all unlimited on the free plan. Cloud transcription is a Pro feature: “If you select Anarlog Cloud or another hosted provider, Anarlog sends the audio or text needed for that request to the selected provider.” The setup steers new users there: “Sign in and choose Anarlog for the quickest setup.”
Anarlog keeps notes, transcripts and recordings on the device by default, and with cloud transcription the audio “travels through our servers to our speech-to-text providers”, deleted “once the job completes.” Listenr deletes meeting audio after a healthy transcription unless you keep it, and never includes audio in a shared copy. Its files are owner-only, which is a filesystem permission rather than encryption.
Anarlog Pro adds end-to-end encrypted sync across devices, and its security page says “Fastrepl cannot read the recovery key”. Listenr has no sync; the library lives on one Mac. Neither holds a certification: Anarlog says it does not claim SOC 2, ISO 27001 or HIPAA, and neither does Listenr.
Who said what, and who writes the notes
Listenr separates the voices on the Mac and names them in three places: the Zoom desktop app through macOS Accessibility (a setting, off by default), Google Meet through the Listenr browser extension, and Microsoft Teams. Elsewhere, including Zoom in a browser, voices stay Speaker 1, Speaker 2. Anarlog offers manual speaker labelling on its free plan and better speaker identification on Pro.
With the language set to auto, the default, Listenr picks the meeting’s language from context such as each contact’s language and the calendar text, and writes the notes in it.
On the notes model, Anarlog offers more. Both take your own API key, a local model and Apple Intelligence. Anarlog can also reuse a Claude, ChatGPT, Grok, GitHub Copilot or Kimi Code subscription with no separate key, which Listenr cannot.
Driving the app from a script or an assistant
Both put a command line tool and a local MCP server on the free plan, reading a database on your own disk. Anarlog’s 11 MCP tools are reads plus a proposal path: an assistant can suggest an edit to a summary or memo, and you decide.
Listenr exposes the whole app. Every user-facing command is declared once and projected onto the menu, the keyboard, the command line and MCP, so all of them are reachable from a script: more than 100 command line verbs, and an MCP server that grows from seven read-only tools to over 150 when you enable control and settings with a flag. An agent can start and stop a recording, pick the language, regenerate notes, edit a transcript, retranscribe a range, deliver to Obsidian or Notion, manage models and change settings.
listenr record start --platform googleMeet --language de
listenr record stop --wait
listenr notes regenerate 8f3a --wait
listenr deliver 8f3a obsidian
listenr do meeting:setTitle --arg meeting=8f3a --arg title="Acme kickoff"
Destructive commands refuse without --yes and name the exact target, and
every outcome is one of seven exit codes, so a script can branch on the result.
Sign-in, connecting a calendar and granting a macOS permission still need a
person: the command returns a code saying so. Account profile, email and
password management stay in the app.
A Mac app, or an app that runs on a Mac
Listenr is written in Swift and SwiftUI for macOS only. Its windows and controls are the system’s own, and it uses Apple’s frameworks directly: Core Audio for call audio, the Neural Engine for on-device speech models, EventKit for the calendar, the Keychain for your keys and Metal for local models. It is signed with a Developer ID, notarized, and updates through Sparkle.
Anarlog’s desktop app is a Tauri build: a Rust core with the interface in a web view, one codebase for macOS, Windows and Linux. That choice is why the Windows and Linux builds exist. If a colleague is on Windows, Anarlog has an answer and Listenr does not.
Where Anarlog is better
Getting meetings into other systems. Anarlog’s free tier has a local API,
webhooks and a hosted cloud API. Listenr has no webhooks and no public REST
API; the nearest thing is listenr watch, which streams events on the Mac
itself.
Reach. Anarlog runs on Apple Silicon and Intel Macs, has Windows and Linux builds in beta, and iOS and Android apps in beta. Listenr requires macOS 15 or later on Apple Silicon and has no phone app. Anarlog also imports audio and transcripts recorded elsewhere; Listenr has no import.
Neither tool announces itself to the other people on the call, so telling them is on you; a bot does it differently.
Pick Listenr when
You are on a Mac, you want transcription on it without changing a setting first, and you want speaker names on Zoom, Google Meet and Teams, plus the participant list and chat from Zoom and Google Meet, to arrive in the notes without labelling anything by hand.
Pick Anarlog when
Take Anarlog if anyone you work with is on Windows or Linux, if a security review wants to read the source, or if you need webhooks and a hosted API. It also syncs between devices with end-to-end encryption, reuses a Claude or ChatGPT subscription, imports audio recorded elsewhere, and costs nothing.
Common questions
Is Anarlog the same app as Hyprnote?
Yes. The app was called Hyprnote, then Char, and took the name Anarlog in May 2026, so older writing about Hyprnote describes this product. Its GitHub repository says the core team is now building char, and that the Anarlog community application remains open source, MIT licensed and maintained.
Is Anarlog free?
Yes, and the free tier is substantial: unlimited on-device transcription, local recordings, your own speech-to-text and model keys, notes, templates, exports, and a local API with a command line tool, MCP server and webhooks. Pro is $14 a month or $140 a year and adds cloud transcription, better speaker identification, end-to-end encrypted sync and sharing.
Does Anarlog transcribe on my device?
It can, without limits on the free plan, using Apple Speech on macOS 26, Parakeet models for 25 European languages, a bundled model, or a whisper.cpp file you supply. Cloud transcription is the paid option, and its setup guide suggests signing in to that first. Listenr’s default is Apple’s on-device engine, and four of its five transcription backends run on the Mac.
Which has better command line and assistant support, Listenr or Anarlog?
It depends on the job. For driving the app, Listenr goes further: every command is a command line verb, and its local MCP server goes from seven read-only tools to over 150 once control and settings are enabled with a flag, so an assistant can start a recording or change a setting. For wiring meetings into other systems, Anarlog goes further, with webhooks and a hosted cloud API that Listenr lacks. Both are free and read a database on your own disk.
Does Anarlog run on Windows, Linux or a phone?
Yes. Anarlog has macOS builds for Apple Silicon and Intel, Windows and Linux builds in beta, and iOS and Android apps in beta through TestFlight and Google Play. Listenr requires macOS 15 or later on Apple Silicon and has no Windows, Linux, web or phone version.
Try it against your own meetings.
The free plan records, transcribes and writes notes with a model on your Mac. No account needed to find out whether the transcript is good enough.