Private AI meeting assistants in 2026: what actually stays on your device

You handle conversations other people expect you to keep private: a client session, a legal intake, a patient call, a board discussion. You also want the notes without spending an hour writing them up afterwards. The problem is that most tools solve the second part by breaking the first. A cloud notetaker records the meeting, then sends the audio to a server you do not control so it can be transcribed and summarised somewhere else.

That is why "private AI meeting assistant" has become a loaded phrase. In 2026 it can mean a genuinely different thing depending on which app you download. EU-hosted cloud. Audio deleted after processing. Local transcription with a cloud summary. Bring your own API key. A fully open-source stack you run yourself. Plain Markdown files you own. Each of these is a real privacy model, and each asks something different of you.

This page compares six private meeting assistants for Mac: Weeve, Jamie, MacWhisper, Anarlog, Meetily, and Slipbox. You get a summary table, a feature comparison, an honest per-tool breakdown with pricing, and a way to pick based on how much of the privacy work you want the product to do for you.

Private AI meeting assistants: a brief overview

• Weeve: the managed local option. Recording, transcription, speaker diarisation, summaries, and storage all run on your Mac, nothing is uploaded, and no bot joins the call. The private path is the default, with no stack to assemble.

• Jamie: connected convenience through an EU-hosted cloud. No bot joins, audio is deleted after transcription, and it runs on Mac, Windows, and mobile, so it fits teams that want cross-platform access and accept off-device processing.

• MacWhisper: a local transcription workstation for Mac. Strong for batch media, file import, and exports, with a one-off licence rather than a subscription.

• Anarlog: open-source and local-first, running on Mac, Windows, and Linux. Notes are stored locally with Markdown export, and you bring your own AI provider. Fully local when paired with a local model.

• Meetily: open-source meeting capture across Mac, Windows, and Linux, with configurable transcription and language-model providers. Genuinely local when you run a local model.

• Slipbox: local capture with a hybrid architecture. Transcription runs on the device; some cloud features are available through your own API keys.

What "private" actually means here

Privacy in this category is not one setting, it is a spectrum of where the meeting content goes.

At one end, a tool keeps the whole workflow on your device: capture, transcription, summary, and storage never touch an external server. At the other end, the recording leaves for a cloud processor and you rely on that vendor's retention and training promises instead. In between sit hybrid tools that process part of the workflow locally and send the rest to the cloud.

A second axis matters just as much: how much you have to configure. A configurable product can stay fully local, but only if you pick the right speech model, language model, provider, and endpoint. That is real control, and it is also real work. One wrong provider setting quietly moves the privacy boundary. The distinction this page keeps coming back to is managed locality: whether the private path is the product's default, or something you have to build and maintain yourself.

How the private AI meeting assistants compare

Currencies differ by vendor and are shown in each vendor's own currency.

1. Weeve: best for keeping the AI and your data on your Mac

You came here because a recording of a confidential conversation should not sit on a server you cannot see. Weeve answers that at the architecture level. You record the microphone and system audio on your Mac, the Mac transcribes and separates speakers, a local model writes the summary, and the content stays in local app storage. There is no upload step, and no bot appears in the call as a visible participant. The honest consent framing still applies: ask the people in the room before you record. On-device processing means the recording never leaves the room, but it does not ask for permission on your behalf.

What sets Weeve apart from the other local tools here is that the private path is the default, not a configuration you assemble. You do not choose a provider matrix or run a self-hosting project. You install it, and recording, diarisation and local summaries work out of the box. The same workflow handles a Zoom call, an in-person meeting across a table, and an old audio file you import later, which matters because sensitive work does not arrive in one format.

Key features

• Recording, transcription, speaker diarisation, and summaries all run on the Mac

• No bot joins the call; Weeve captures the Mac's own audio

• Works with Zoom, Google Meet, and Microsoft Teams with no integration to install

• Imports audio and video files, including MP4, and transcribes them locally

• Projects and templates to organise notes, plus everyday dictation

• Local AI models, with announced support for Gemma 4 summaries

Best use cases

• A therapist running 20 client sessions a week who cannot put session audio in the cloud

• A notary or small legal practice keeping privileged conversations on the device

• A consultant who moves between video calls and in-person client meetings and wants one local workflow for both

Pricing

• Free Starter plan: EUR 0, 10 recordings a month, 2 projects

• Pro: EUR 12,99 a month excl. VAT, unlimited recordings and projects

Pros

• Both the AI and the recording stay on the Mac, so there is no cloud retention promise to audit

• No bot in the participant list, which removes an awkward moment with clients

• No upload step means no upload step to fail mid-meeting

• Built by two founders in Amsterdam, backed by Achmea Impact Ventures, a useful trust credential for risk-sensitive buyers

Cons

• Mac only today: no Windows and no mobile app

• Summaries arrive after the call rather than live, and speaker labels can need a correction

• You own device security and backups, which is the trade-off for keeping the data local

If you want to try the private path first, you can download Weeve for Mac and start on the free Starter plan.

2. Jamie: best for cross-platform teams that accept cloud processing

Jamie is the connected-convenience option. It records without adding a bot to the call, then processes the audio in an EU-hosted cloud and returns structured notes, transcripts, and action items in 99+ languages. Its own site states that meeting audio is deleted after transcription and that customer data is not used to train models, with GDPR compliance and EU data residency. For an organisation that has decided the cloud is acceptable as long as it is EU-hosted and does not train on their content, that is a defensible position.

Jamie's real advantage over the local tools on this page is reach. It runs on Mac, Windows, iPhone, and Android, so the notes follow you across devices. The switch reason, if privacy is your priority, is equally plain: the recording is processed off your device on someone else's servers, and you are trusting a retention and training policy rather than an architecture where the audio never leaves.

Key features

• No bot joins; Jamie captures audio on the device

• EU-hosted processing with stated EU data residency

• Transcripts, summaries, and action items in 99+ languages

• Apps for Mac, Windows, iPhone, and Android

Best use cases

• A distributed team on mixed hardware that needs notes on every device

• Users who want mobile capture as well as desktop

Pricing

• Free: EUR 0, 10 meetings a month, 30-minute limit

• Plus: EUR 21 a month billed annually, 20 meetings a month

• Pro: EUR 39 a month billed annually, unlimited meetings

Pros

• Genuine cross-platform and mobile support

• Audio deleted after transcription, per Jamie's own policy

• Stated policy of not training models on customer data

Cons

• The recording is processed in the cloud, off your device

• Privacy rests on a retention promise, not on the audio staying local

3. MacWhisper: best for transcribing files and media on a Mac

MacWhisper is a local transcription workstation rather than a meeting-memory product. It runs Whisper-family models on the Mac to transcribe audio and video without the data leaving the machine, and it is strong at the jobs a meeting tool treats as secondary: batch transcription of many files, drag-and-drop import, speaker recognition, and a wide range of exports including SRT, VTT, TXT, Markdown, PDF, and DOCX. It can also record calls on Zoom, Teams, and similar apps without a bot, and it includes system-wide dictation.

Where it differs from Weeve is scope. MacWhisper is built around transcribing content, not around projects, templates, and organised meeting memory. If your main need is turning a stack of recordings or media files into accurate local text, it is excellent, and the one-off licence is a genuinely different pricing model from a subscription.

Key features

• Local Whisper-based transcription, nothing uploaded

• Batch transcription and drag-and-drop file import

• Records Zoom, Teams, and other calls without a bot

• Speaker recognition and exports to SRT, VTT, TXT, MD, PDF, and DOCX

• System-wide dictation

Best use cases

• A journalist or researcher transcribing interviews and media files locally

• Anyone who prefers a one-off purchase to a subscription

Pricing

• Free: EUR 0, forever

• Pro: EUR 64 one-off licence with lifetime updates

Pros

• Fully local transcription on the Mac

• Strong batch and export tooling

• One-time cost rather than a recurring fee

Cons

• Built for transcription, not organised meeting memory with projects

• The MacWhisper app is Mac only; the same maker's iOS transcription app is a separate product

4. Anarlog: best for an open-source, own-your-data workflow

Anarlog is the open-source, own-your-data option. It records meetings without a bot and stores your notes locally on your device, with Markdown export when you want portable files. It runs on Mac, Windows, and Linux. It is bring-your-own-AI: you connect the provider you want for summaries, from OpenAI, Anthropic, Gemini, or OpenRouter to local engines like Ollama and LM Studio. Paired with a local engine, both transcription and summarisation run entirely on your machine.

That openness is the appeal and the cost. Anarlog gives you portability and control that a closed app cannot, but it also asks you to choose and wire up your own AI stack, and the privacy boundary depends on that choice. This is the same genuinely local capability as Weeve, offered as an open toolkit rather than a managed default.

Key features

• Records without a bot, notes stored locally with Markdown export

• Bring-your-own AI: OpenAI, Anthropic, Gemini, OpenRouter, Ollama, or LM Studio

• Fully local when paired with a local model

• Runs on Mac, Windows, and Linux

Best use cases

• A developer or power user who wants notes as portable, version-controlled files

• Anyone committed to a plain-text, own-your-data workflow

Pricing

• Free forever for local transcription and bring-your-own-key use

• Pro: USD 15 a month, or USD 150 a year

Pros

• Open source and genuinely local with a local model

• Notes export to portable Markdown, and the data stays on your device

• Flexible choice of AI provider

Cons

• You assemble and maintain the AI stack yourself

• The privacy boundary moves with the provider you choose

5. Meetily: best for technical users who want full control

Meetily is an open-source, self-hostable meeting assistant under an MIT licence. Its README states that all processing is done on your machine and that no data leaves your computer, and it runs on macOS, Windows, and Linux. It is genuinely local, not local with an asterisk: run it with Ollama and both transcription and summaries stay on the device. It also lets you point it at other providers, including Claude, Groq, OpenRouter, or a custom OpenAI-compatible endpoint, which is what makes it a fit for organisations with their own AI infrastructure.

The contrast with Weeve is philosophy, not privacy. Meetily is open, configurable, and technical, aimed at people who want to choose their own models and are comfortable installing and maintaining the setup. Weeve is the opinionated, no-setup counterpart for a non-technical professional who wants the private path to work out of the box. Meetily is also the only tool here besides Jamie that runs on Windows and Linux, which is a real advantage if you are not on a Mac.

Key features

• Open-source, MIT-licensed, self-hostable

• Fully local processing with a local model such as Ollama

• Configurable transcription and language-model providers

• Runs on macOS, Windows, and Linux

Best use cases

• A technical team that wants to self-host and pick its own models

• Non-Mac users who still want a local-capable assistant

Pricing

• Community edition: free and open-source

• Meetily Pro: paid tier for enhanced accuracy and features

Pros

• Genuinely local and open-source

• Cross-platform, including Windows and Linux

• Provider flexibility for custom AI infrastructure

Cons

• Setup and maintenance expect technical comfort

• No managed, no-setup experience

6. Slipbox: best for live meeting help plus local capture

Slipbox pairs local capture with a hybrid architecture. It records system audio and your microphone on the device with no bot, and transcription and speaker identification run locally. On top of that it adds cloud features: a live "ask anything" assistant during the meeting and a natural-language search across your meeting history. Those cloud features are opt-in, and Slipbox offers a bring-your-own-key option for the model provider.

That hybrid design is the honest distinction from Weeve. Slipbox's local capture is real, but the intelligence layer reaches into the cloud when you turn it on, whereas fully on-device is Weeve's default with nothing to enable. If live in-meeting assistance and semantic memory are worth a partial cloud path to you, Slipbox is a strong option, and it runs on both Mac and iPhone.

Key features

• Local recording and transcription with no bot

• Live "ask anything" assistance during meetings

• Natural-language search across meeting history

• Bring-your-own-key option for the cloud AI features

Best use cases

• Someone who wants live prompts and answers mid-meeting

• A Mac user who also wants capture on iPhone

Pricing

• Hobby: free

• Pro: USD 99 a year

• Enterprise: custom pricing

Pros

• Local capture with no bot in the call

• Live assistance and semantic search across past meetings

• Mac and iPhone support

Cons

• The intelligence layer is hybrid, so parts of the workflow can touch the cloud

• Cloud features depend on your own provider keys

Which private AI meeting assistant is right for you?

If your conversations are confidential by profession

Choose Weeve. If you are a therapist, notary, clinician, consultant, or work in IT and security, the sessions and privileged calls stay on the Mac, and there is no cloud processor to explain to a client or a regulator. You get the managed local workflow without building it, which is the point when your day is meetings, not model configuration. It is worth reading why a private note-taker should stay off the cloud before you decide. And if a genuinely free tier is the deciding factor, our roundup of the best free meeting recorder apps covers that angle.

If you need a mobile app or Windows support

Look at Jamie or Meetily. Jamie gives you cross-platform apps and mobile capture through its EU-hosted cloud, which suits a mixed-hardware team that accepts off-device processing. Meetily runs on Windows and Linux as well as Mac and stays local with the right model, if you are comfortable self-hosting. Weeve is Mac only today, so for either of these needs it is not your tool.

If you want open source and to own your data

Choose Anarlog or Meetily. Anarlog is open-source and cross-platform, with Markdown export and your own choice of AI provider, and Meetily is a self-hostable, MIT-licensed stack you can point at your own models. Both are genuinely local with a local engine. The trade-off is that you assemble and maintain the AI setup, where Weeve does that for you.

If you mainly transcribe files and media

Choose MacWhisper. For batch-transcribing interviews, lectures, or a backlog of recordings locally, with strong exports and a one-off licence, it is the most direct fit. It is a transcription workstation rather than a meeting-memory tool, so pair it with something else if you also need organised notes.

FAQ

What is a private AI meeting assistant? A private AI meeting assistant records and summarises your meetings while limiting where the audio and transcript go. The strictest versions keep everything on your device, so the recording never reaches an external server. Others process in the cloud but promise to delete audio or avoid training on it. The privacy model varies by tool, so check where each one actually processes and stores your content.

Which AI meeting assistant keeps everything on your device? Weeve keeps recording, transcription, speaker diarisation, and summaries all on your Mac, with no upload step. MacWhisper is also fully local for transcription, and Anarlog and Meetily are genuinely local when paired with a local model such as Ollama. The difference is that Weeve makes the local path the default with no setup, while Anarlog and Meetily ask you to assemble the stack.

Does Jamie store your meeting recordings? Jamie processes meeting audio in an EU-hosted cloud and states on its site that the audio is deleted after transcription, as of July 2026. It also states that customer data is not used to train models. This is a cloud model with a retention promise rather than an on-device architecture, so the recording does leave your machine to be processed.

Is there an AI notetaker that works without a bot joining the call? Yes. Every tool on this page captures audio without adding a visible bot to the meeting, including Weeve, Jamie, MacWhisper, Anarlog, Meetily, and Slipbox. Weeve does it by capturing the Mac's own system audio, so it works with Zoom, Google Meet, and Microsoft Teams without any integration or participant appearing in the call.

Can Weeve record in-person meetings, not just video calls? Yes. Because Weeve captures the Mac's microphone and system audio directly, it records an in-person conversation across a table the same way it records a video call. It also imports existing audio and video files and transcribes them locally, so a call, a meeting, and an old recording all go through one local workflow.

Is Weeve or Slipbox more private by default? Weeve is fully on-device by default, while Slipbox is hybrid. Both capture and transcribe locally with no bot, but Slipbox's live assistant and cross-meeting search are opt-in cloud features, whereas Weeve keeps summaries and search on the Mac with nothing to turn on. If you want zero cloud in the workflow out of the box, Weeve is the stricter default; if you want live in-meeting help, Slipbox's hybrid path may be worth it.

Is Meetily less private than Weeve because it is open source? No. Meetily is genuinely local: run with a local model, all processing stays on your machine, and it is open-source under an MIT licence. The difference from Weeve is philosophy and effort, not privacy. Meetily is configurable and technical and expects you to install and maintain it, while Weeve is an opinionated, no-setup Mac experience. If you want full control and can self-host, Meetily is a strong choice.

What is the best private AI meeting assistant for a Mac? For a privacy-sensitive Mac professional who wants the private path without maintaining it, Weeve is the strongest managed choice, because the AI and the recording both stay on the Mac by default. If you want an open toolkit and portable files, Anarlog or Meetily fit better; if you need Windows or mobile, Jamie or Meetily are the honest answer. Check setup and formats on Weeve support if you are weighing the switch.

Private should not mean local only if you configure it correctly. It should mean the normal path keeps the conversation on your Mac.