Anriss
Telling an AI “make the pricing section nicer” is a guess — it burns tokens and often changes the wrong thing. Click the exact element, say what you mean, and your agent gets it right the first time.
Click the elements you mean. Each is captured with everything your agent needs to find it.
Write one comment. Reference the elements right inside your sentence — no jargon, no selectors.
Your agent applies the change and marks it resolved — live, right in front of you.
Every round of “no, the other button” is tokens spent and time lost. Pointing removes the guesswork — your agent gets the exact element and the exact ask.
Words, back-and-forth, hoping the AI finds the right spot.
“Make the pricing section better… no, the card on the left… the button.”
3–4 attempts · tokens burned on guesses
One click captures the exact element and its context.
“Center 1 and drop the toggle.”
1 attempt · fewer tokens, right the first time
AI coding agents bill by the token, and heavy usage keeps getting pricier. A wrong guess is never free — you pay for the mistake and the correction.
“make the pricing section nicer… no, the other card… the button.”
“fix 1” — the exact element you clicked
The exact element, captured for you. Your agent stops guessing where you mean.
Precise context means fewer retries — and fewer wasted tokens on your AI bill.
Everything stays on your machine. No cloud, no account, nothing to sign up for.
Claude Code, Codex, opencode — bring the coding agent you already use. No switching.
One line drops it into any site. Nothing to rebuild, no framework required.
Click, type a sentence, done. If you can leave a sticky note, you can use Anriss.
Anriss
If it renders in a browser, Anriss annotates it — nothing to rebuild, no framework to adopt.
Other tools each do a slice of this — but tie you to their cloud, their app, or one framework. Here’s who does what.
Anriss
Point at any app, in your own browser. Your note goes straight to the agent you already use and the fix comes back to the page. Your data never leaves your machine. Free and open source.
Annotate inside OpenAI’s own app. Deep — but it only feeds Codex, only in their browser.
Point-and-comment bug reports collected onto a cloud task board. Built for triage, not your build loop.
Website feedback with screenshots, routed to your issue tracker — all through their cloud.
AI design agents that generate and edit UI from prompts — for creating designs, not annotating your live app.
Describe a UI and get a rendered preview in chat. Great for mockups — but it isn’t pointing at your running app.
In-app comment widgets — often locked to one framework, or leaving notes your agent can’t act on.
An Anriss is the fine line a machinist scribes onto a workpiece to mark exactly where to cut — before a single chip is removed. Mark the precise spot first; then the work happens.
That’s the whole idea. You scribe the change onto the page — your agent does the cutting.
Your annotations never leave your machine — no cloud, no account. Nothing to leak, nothing to put in a privacy policy.
Try it on a live page right now — no install, no sign-up.
Open the live demo