Skip to main content
PromptQuorumBuilt for humans. Structured for AI.
Home/Power Local LLM
Productivity & Knowledge Tools

Dayflow Review: Open-Source Mac Work Journal with Local AI

·7 min readBy · Founder of PromptQuorum · Discovery engine for open-weight & open-source AI

Dayflow is a free, MIT-licensed Mac app that records your screen at low frequency and uses AI to turn it into a timeline, and its analysis can run entirely on your own Mac through Ollama or LM Studio. Cloud analysis is a choice, not a requirement. The newest release is version 2.9.3 from 13 November 2026. This review rests on the maker's website and GitHub repository, not on hands-on testing.

Key Takeaways

  • Local claim: capture, recordings and the database stay on the Mac; analysis is local only if you connect Ollama or LM Studio.
  • License and maker: MIT, open code, published by Teleport Labs Inc. per the privacy page.
  • Price: free with a local model or your own key; managed AI is the paid Pro plan.
  • Maintenance: version 2.9.3 was released on 13 November 2026 and the repository was updated the same day, so the project is actively developed.
  • Gaps: macOS 14 or newer only, no bundled model, and cloud paths send screenshots to a provider.

📍 In One Sentence

Dayflow is a free, open-source Mac app that, per its README, turns low-frequency screenshots into a work timeline and can run the AI analysis entirely on your own Mac.

💬 In Plain Terms

Think of an automatic diary of your workday: instead of starting timers, you let the app watch your screen every so often and write down what you were doing.

Read in:
🇺🇸en🇩🇪de🇫🇷fr🇯🇵ja🇨🇳zh🇪🇸es🇧🇷pt🇸🇦ar🇰🇷ko

Dayflow is a free, MIT-licensed Mac app that takes low-frequency screenshots of your screen and turns them into a timeline of what you worked on, with daily and weekly summaries and a chat over your own timeline. This review covers what the maker's website and GitHub repository say, which parts can run on your own Mac through Ollama or LM Studio, where screenshots leave the machine, and how Dayflow compares with other tools in the directory.

Dayflow Review: Open-Source Mac Work Journal with Local AI

Get Dayflow

Dayflow installs like any Mac app, so no terminal is needed. Download the disk image from the official site or the GitHub release page, open it, drag Dayflow into Applications and grant the screen recording permission when macOS asks.

Platform
Get It
MacDownload for Mac — Dayflow.dmg, macOS 14 or newer; also brew install --cask dayflow
Release notesGitHub Releases — the 2.9.3 release has no written notes
Websitedayflow.so
Privacy policyDayflow privacy policy
Source codegithub.com/JerryZLiu/Dayflow — MIT license

The homepage advertises a Windows version as coming; no Windows or Linux build was found.

Dayflow at a Glance

Attribute
Dayflow
PlatformmacOS 14 or newer
PriceFree; paid Pro plan
LicenseMIT, open source
RunsCapture on the Mac; analysis local or cloud
Works with local LLMsYes, via Ollama or LM Studio
Voice input / outputNot stated
Persistent memoryTimeline database on your Mac
Tool use / actionsNot stated

Attributes follow the personal-assistants comparison used in the Local LLM Software Directory. A missing statement on the site means unknown, not unsupported.

What Dayflow Is

Dayflow is an automatic work journal, not a manual timer. It takes low-frequency screenshots, has an AI model describe what you were doing, and groups the result into activity cards on a timeline, so a block of Cursor or Chrome shows up as, say, debugging or research rather than as an app name.

The homepage names Teleport Labs Inc. as the company, and the GitHub repository is JerryZLiu/Dayflow. The pricing page also lists an Enterprise plan with SSO and data-retention controls; this review covers the Mac app only.

This review draws on the website, the privacy policy and the GitHub repository. It involved no hands-on testing, so the quality of summaries, battery cost and stability are not rated here.

How to Get Started

Setup is an install, a permission and an AI choice. The steps follow the README.

  1. 1
    Install the app
    Why it matters: Download Dayflow.dmg from the [GitHub release page](https://github.com/JerryZLiu/Dayflow/releases/latest), drag it into Applications, or run `brew install --cask dayflow`.
  2. 2
    Grant the permission
    Why it matters: Allow macOS Screen & System Audio Recording when prompted; without it Dayflow cannot capture anything.
  3. 3
    Choose how analysis runs
    Why it matters: Pick a local model through Ollama or LM Studio, your own Gemini key, or ChatGPT or Claude through their local CLI tools; Pro is the managed alternative.
  4. 4
    Let it run, then review
    Why it matters: Work as usual; open the timeline, the daily view or the weekly review later, or export a date range as Markdown.

Features

The feature set centres on the timeline. Everything below comes from the README and the website.

  • Automatic timeline. Screen chunks become activity cards with context, such as what you did in an editor or browser.
  • Daily standup. An activity grid plus yesterday's highlights, today's priorities and blockers.
  • Weekly review. Focus patterns, categories and app usage across the week, with distraction tracking.
  • Chat with your journal. Natural-language questions answered from your own timeline.
  • Export and cleanup. Markdown export for any date range and automatic purging of old recordings against a storage limit.

Local vs. Cloud: Where Screenshots Go

Dayflow offers three analysis paths, and only the first keeps screenshots on the Mac. The privacy policy lists Local processing, where screenshots never leave the machine; Bring your own AI provider, where screenshots go straight to the provider you configure; and Dayflow cloud for Pro, where screenshots are processed on the maker's servers and deleted afterward. It labels the first two "default" and names no single default.

Local processing runs through Ollama or LM Studio, which you install separately. The README also lists ChatGPT and Claude through their local command-line tools; those tools call the vendors' cloud models, so this path is not on-device inference. The README and the pricing page place recordings and the database on the Mac in every mode.

Privacy, License and Cost

The privacy policy says Dayflow does not sell your data, does not store your screenshots and does not let its AI providers train on your data. It also says anonymous analytics (app launches, feature usage, crash reports) can be switched off in settings, and that basic account information is kept for subscribers. The company is US-based. The code is open and PromptQuorum did not examine network traffic, so these statements are the maker's own.

The license is MIT. The free plan covers capture, summaries and your own provider keys or a local model; Pro costs $20 a month for managed AI, with an annual offer shown in the app, and Enterprise is priced on request.

Trade-Offs: Benefits vs. Limitations

Analysis on your own Mac

What it means in real use:
Per the README, with Ollama or LM Studio, screenshots stay on the machine.
Limitation / caveat:
You install and run the model yourself; local inference is demanding on GPU and battery per the homepage.

Free and open source

What it means in real use:
MIT code, free plan with your own key or a local model.
Limitation / caveat:
Managed AI needs the paid Pro plan.

Automatic timeline

What it means in real use:
No timers or manual notes; summaries describe what you did.
Limitation / caveat:
Summary quality depends on the model you choose; PromptQuorum did not measure it.

Flexible AI choice

What it means in real use:
Switch between local, your own key and Pro later.
Limitation / caveat:
Cloud paths send screenshots to a provider.

Active development

What it means in real use:
Version 2.9.3 shipped on 13 November 2026.
Limitation / caveat:
Mac only, macOS 14 or newer; Windows is only announced.

Dayflow vs. Alternatives

These are the closest tools in the directory: another screen-memory app, a self-hosted second brain and an on-device file search.

Dayflow

Platforms:
Mac
Price and license:
Free, Pro paid, MIT
Role:
Automatic work journal
Key difference:
Open source; local models via Ollama or LM Studio

Perfect Memory AI

Platforms:
Mac, Windows
Price and license:
Freemium, closed source
Role:
Searchable screen and meeting memory
Key difference:
Built-in AI uses the maker's server unless you connect a local one, per its privacy policy
Articles about Perfect Memory AI (1)

Khoj

Platforms:
Mac, Windows, Linux, web
Price and license:
Free, AGPL 3.0
Role:
AI second brain over your documents
Key difference:
Works on notes and files you provide, not on screen activity
Articles about Khoj (4)

Also mentioned in:

+4 more not shown

Platform, price, and feature details for third-party apps change frequently. Verify current specifics on each project's own page before deciding.

Compare in directory

Who Should Use Dayflow

  • Mac users who bill or report their time. The timeline and Markdown export replace manual notes.
  • Developers who write standups. The daily view drafts yesterday's highlights, today's tasks and blockers.
  • People who keep screen data private. A local model keeps analysis on the Mac, per the README.
  • Open-source users. The MIT code is public and the app is free.

Who Should Not Use Dayflow

  • Windows and Linux users. Only a Mac app exists; Windows is announced.
  • Macs older than macOS 14. The README requires macOS 14 or newer.
  • Anyone unwilling to run a local model. Without Ollama or LM Studio, analysis uses a cloud provider.
  • Anyone who cannot record their screen. Employer or client rules may forbid periodic screenshots.

Frequently Asked Questions

Does Dayflow work offline?

With a local model in Ollama or LM Studio, analysis needs no internet connection, per the README; updates, account sign-in and cloud modes do. PromptQuorum did not test it offline.

Does Dayflow record my screen continuously?

The privacy policy says it captures screenshots at low frequency, and the homepage says the recorder uses around 100 MB of RAM and under 1% CPU. Those are the maker's figures; PromptQuorum did not measure them.

Can I use ChatGPT or Claude instead of a local model?

Yes, through their local command-line tools (Codex CLI and Claude Code) or with your own Gemini, OpenAI or Anthropic key. These paths use cloud models, so screenshots reach the provider.

Where are my recordings stored?

In the Dayflow folder inside Application Support on your Mac, per the README. Activity Data is stored on the maker's servers only if you subscribe to Pro for cross-device sync.

Is there a Windows version?

Not yet: the homepage says Dayflow for Windows is coming and offers a waitlist, and no Windows build was found.

Verdict

Dayflow is a good fit if you use a Mac, want an automatic and open-source record of your workday, and are ready to run Ollama or LM Studio so the analysis stays on your machine. It has an MIT license, a free tier and a release on 13 November 2026.

The caveats are scope and verifiability: Mac only, a separate local-model setup, cloud paths that send screenshots away, and privacy and accuracy claims that PromptQuorum could not check. Try the free local setup before considering Pro.

Sources

← Back to Power Local LLM

MCP