For macOS
What people ask you becomes a task on its own.
Radar reads your screen, your Slack and your calls. When someone asks you for something, it lands on your list with who asked and where it came from, and when the answer is in your code, the reply is already written.
Reads what crosses your screen
Slack, WhatsApp, email, docs. Even what shows up for a few seconds becomes a task suggestion, with who asked and when it's due.
Never misses a mention
Reads your DMs, mentions and whole threads, screenshots included. Real requests become tasks; bot alerts are skipped.
Answers from your code
Connect GitHub and questions about your product come with where it lives in the code and a reply ready to send.
Records calls and in-person talks
Starts when a Meet, Zoom, Teams or huddle begins. You get the summary, the decisions and to-dos with owners.
Ask about your work
"What is overdue?", "What did Maya ask me this week?". Answers point to the task or call they came from.
Your Mac, your AI
No Radar servers. Run it on a free local model, or on Claude, ChatGPT or Gemini with your key. It never sends a message for you. How it stays private →
The question arrives with the answer
Someone asks how something works. Radar reads the conversation, searches your repositories, explains where the behaviour lives with links to the exact lines, and writes the reply in their language.
Edit it if you want, then Save as Slack draft: it waits in the right thread for you to send.
New task or new details, clearly
When people keep talking about something you already track, Radar doesn't open another task. It shows exactly what gets added to the one you have.
Summary, decisions and who does what
When the call ends, Radar writes what was decided and turns each request into a to-do with its owner, the exact words and the due date. Accept it and it lands on your list.
Ask what you're forgetting
Your tasks, suggestions, calls and screen readings, searchable in plain words. Every answer cites its source; click it to open that exact task or call.
You always know what the AI costs
See today's and this month's spend in dollars and tokens, split by screen, calls, chat and code. Or pick the local model and pay nothing.
Your work stays on your Mac
Radar has no servers. There is no Radar account and no Radar cloud: everything it reads and writes lives on your Mac, and you choose who, if anyone, does the thinking.
Run it fully local
Pick the local model and nothing leaves your Mac: Qwen3 4B runs on your own chip, for free, with every feature working. Prefer a bigger model? Use your own Claude, ChatGPT or Gemini key, or the Claude Code and OpenCode you are already signed in to. Data goes straight from your Mac to the provider you picked, with no stop in between.
- Local databaseTasks, transcripts, chats and settings live in a database on your Mac. Your API keys and tokens too.
- Text, not recordingsThe screen is read as text, no screenshots kept. Call audio is deleted once it is transcribed.
- Apps it never readsPassword managers, System Settings and any app you add to Ignored apps are skipped.
- It never acts for youRadar never sends a message, posts, reacts or changes anything in Slack or GitHub. It only saves drafts you ask for.
- Read-only accessSlack and GitHub are connected with read permissions, plus drafts. You pick which repositories it may read.
- Signed updatesEvery update is signed and checked before it installs, and you can see exactly what the AI cost you.
How to install
- Open Radar.dmg. If macOS says it can't verify it, click Done, go to System Settings → Privacy & Security, scroll down and click Open Anyway.
- Drag Radar to Applications and open it. If it is blocked again, do the same Open Anyway step.
- Allow Screen Recording and Microphone when it asks.
Still blocked? Drag Radar to Applications, then paste this in Terminal: xattr -dr com.apple.quarantine /Applications/Radar.app
Recording calls needs macOS 15. Transcribing them needs macOS 26.