🎬Summary
Paste a link, get the breakdown.
A YouTube link. A wall of text. A question. An error message. Handler knows what each one is and what it's for, before you've written a single prompt.
Your tone, your sources, your lines that never get crossed. Every decision runs through your rules before anything happens.
Choose a workflow from your library, with your rules controlling when it can run.
The steps, the tools, the destinations: laid out before anything happens. You approve the plan, not each click.
Each step runs in the open, in order, with tools you put in its toolbox. No surprises, no side quests.
Watch the words stream in, usually in under a second.Preview the steps, then follow the run as it progresses. Recovery depends on the action; external changes may need to be undone in the destination app.
Copy a sample, choose a workflow, then paste it here to try a prepared example. No install or account needed.
John Smith · 42 Main Street, Springfield, 00000 · (555) 555-0142 · john.smith@example.com
The wheel appears at your cursor. Click a workflow, then paste the sample in the next field. Or choose from this keyboard-friendly list.
Only the selected example is accepted. Pasted text stays in this page; nothing is sent to a model.
These are prepared examples, not live AI responses. In the app, processing uses your chosen local model or hosted provider. Hosted providers receive task content. Form examples use fictional details.
Ships with a default workflow library of templates (Summarize it, Professional rewrite, Answer the question), more in Settings, and you can define your own.
Paste a link, get the breakdown.
Typos and grammar, fixed.
Rough text in, polished out.
Ask a question and read the answer.
The five-line version of anything.
The cause and the fix, in plain English.
Every line, explained.
Same code, another language.
Paste the diff, get the message.
Title, repro, severity, ready to file.
Resolve corrected dates, identify commitments, and keep open questions visible.
From your notes, in your voice.
No, without burning the bridge.
Drafted for your approval.
The traps, in plain English.
Your experience, matched to the posting.
How it was meant, in their language.
Ask in English, get the answer.
Dedupe, fix dates, reformat.
A clean renaming scheme.
Dates and places, formatted.
The list, by aisle.
One word, your whole day.
Speak, get clean text.
Three options, one briefing.
The numbers, explained.
Every field, decoded.
Free and private by design. It works only on what you give it, follows only the rules you set, and does nothing until you tell it to: every step visible, everything reversible.
What Clippy should have been, without the nagging.
Copy anything and the workflows that apply light up on your deck. Agent Deck is an extended control and feedback surface; Handler is what drives it. Neither needs the other, and both ship behind the same download.
Yes, and not just during the alpha. Set up your own keys or run fully local and it's 100% free, forever, on your machine. During the alpha, even our hosted convenience is free; after that it's optional, with posted rates and no surprises.
Run it locally and the answer is simple: we never see your content, so there is nothing we could do with it. Choose hosted inference and we still never log or record what you handle or say; we keep only metering facts (which model, request size, cost) so your runway is honest. Telemetry, when present, is anonymized and never includes content: during the free alpha it's on by default (one setting in the app switches it off), and once anything is paid it becomes strictly opt-in, with a 10% discount for opting in.
Because what you give Handler and what you say is yours. Local mode means nothing egresses: no account, no keys, no telemetry, works offline. The hosted option exists for one reason: convenience, when you'd rather not manage keys or models. Your call, per machine, changeable any time.
Only at the level you choose. The conservative mode shows a plan and waits for approval. Optional autopilot shows a countdown for each step and stays pausable and abortable while it works. Screen understanding is also invoked by you, never always watching.
Windows and Linux are the supported alpha platforms today. macOS is planned for a later stage, after we have a real Mac test and release path.
Paid tiers are optional hosted inference. See exactly what each would get you.
© The Julian Utility Company · Utility Source License 1.0: free for personal use, commercial use requires a license. Questions? support@julianutility.com · Terms · Privacy · Refunds