I sell that waiting line of screen, and pay them 70% of it. One line of text in your status bar, only while the model is working. Cash out in crypto from $5.
Prefer reading it before running it? The installer is eighty lines. Uninstall is ~/.tick/uninstall.sh, and it puts your settings back byte for byte.
A tick means it works today, on a machine we tested. Nothing here is a plan.
Where do you use Claude Code?
| Status | ||
|---|---|---|
| Claude Code in a terminal | Works | Any terminal: iTerm2, Apple Terminal, Ghostty, Kitty, WezTerm |
| Claude Code in the VS Code or Cursor terminal | Works | Same CLI, same hook |
| The Claude desktop app | No | It does not run a status line command. We installed the client on one and watched it never get called |
| VS Code / Cursor side panel | Not yet | An extension is the next thing we build, after payouts |
| Codex, OpenCode, Cursor's own agent | No | Codex exposes no slot for outside text. We will not patch one in |
Three steps, and you only do the first one.
One command. It adds a single line to your Claude Code settings and nothing else. Removing it puts that file back exactly as it was.
A short sponsored line shows while the model is thinking. Nothing pops up, nothing moves, nothing interrupts you. When no ad is running you get the line for yourself.
From $5 you withdraw to a crypto wallet. No documents, no bank, no waiting to be approved — just an address.
It’s one line of text that only shows while you’re already waiting. No images, no popups, nothing that moves. And you can put your own line there instead — a link to your work, anything you like.
We never touch your Claude Code install, and your code never leaves your machine. The client is open source, so you can read every line of it before you run it. That’s the whole reason to pick us over the one that patches your editor.
This is the whole reason to pick us, so here it is without hedging. A test in our CI drives the real scripts and fails the build if any forbidden string shows up in an outgoing request.
There is no mode that trades your conversation for a higher rate. We did not build one and we are not going to: it would cost the only thing that makes this worth installing.
We are second to this idea, so the honest thing is to be specific about what is actually different rather than claim to be new.
| TICK | The incumbent | |
|---|---|---|
| Your install | Untouched. One field added to settings.json, backed up first | Patches Claude Code; reverting is a separate command you have to remember |
| Revenue share | 70% of gross | ~50% of net |
| Getting paid | Crypto from $5. A wallet address | Stripe Connect from a higher threshold. Full KYC, supported countries only |
| Your conversation | Stays local. No paid exception | An opt-in mode pays more for sending it to their servers |
| Client source | Public. Checksums published | Closed |
| Where it runs | Claude Code in a terminal, any terminal. Not the desktop app, not an editor panel | VS Code and Cursor panels, plus the Claude Code CLI |
That last row is theirs, not ours. We would rather print it than have you find out after installing.
These are the rules. What you actually earn depends on how much advertising is running, and nobody can honestly promise you that in advance.
| Your cut | 70% of what an advertiser pays. Before our costs, not after |
| What counts | Five seconds of the line being on screen while the model is actually working. A terminal left open on a desk earns nothing |
| Roughly | $1.40 for every thousand of those. Two hours of real work a day puts it in the tens of dollars a month, if the inventory is sold |
| Cashing out | From $5, to a crypto wallet. No documents, no bank |
| The wait | 30 days from earning to withdrawing. That window is how fraud gets caught before it gets paid |
| Quiet days | When nothing is sold, the line is yours and earns nothing |
What we will not do is quote you a monthly figure. Our own working model says a few dollars a month for someone who lives in Claude Code, and a model is worth what you paid for it. The incumbent promised its users the tool would pay for their subscription, and now gets written about through the gap between that promise and the complaints. When we have real numbers from real sessions we will publish the median — not the average, which one heavy user drags upward and nobody else recognises.
Advertisers pay us in crypto too, invoiced by a person. Nothing is automated in either direction yet, which is slower and lets us read every campaign before it runs.
One line of text, in front of developers, at the moment they are waiting rather than reading. Sixty characters, no images, no scripts.
| Unit | 1,000 impressions. One impression is 5 seconds of continuous display in an active session |
| Floor | $1.00 per 1,000 |
| Clicks | Counted by our own redirect, with your UTM tags preserved. We never bill a click twice |
| Invalid traffic | Filtered impressions are not billed, and the filtered share is reported to you |
| Campaigns | Set up by hand right now, by a person, with youself-serve soon |
| Paying us | Crypto — USDC on Solana, Base or TRON. Invoiced and settled by a person, so you talk to someone before anything runs |
| Moderation | Every campaign is read by a human before it serves. No exceptions, and some categories are declined outright |
We have not published a click-through rate yet, because we do not have an honest one yet. When we do, it goes on this page with the method next to it — including the part where terminals without hyperlink support make it an undercount. Want in early? Write to us.
The script has a 30 millisecond budget and currently runs in about 23. It makes no network calls at all — a background process does that, separately. If the script ever fails, the line goes blank and nothing else changes.
Claude Code hides most of them whenever any status line is configured, ours included: esc to interrupt, ? for shortcuts and the dictation hint. That is upstream behaviour, not something we add. Escape still interrupts; only the reminder is gone.
Because an advertiser is paying for attention, and a terminal left open on a desk has none. We count time only while the model is doing measurable work. Faking that means burning real tokens, which costs more than it pays.
We are not going to put a number here. Our own projection says a few dollars a month for a developer who uses Claude Code heavily, and projections are worth what you paid for them. When we have real medians from real sessions, we will publish the median rather than the average.
No. The desktop app does not run a status line command — we installed the client on one and watched it never get called. TICK needs Claude Code in a terminal: iTerm2, Apple Terminal, the VS Code or Cursor integrated terminal, anything. If that is not how you work, this is not for you yet, and we would rather say so than take the install.
Not today. Codex only lets its status line show its own built-in fields; there is no hook for external text, and we will not patch it to make one. OpenCode is reachable and planned. Cursor exposes nothing. We will put a tick in that table on the day it works, not before.
Yes, and unsold inventory goes to you by default. Put own_line in ~/.tick/config.json and it shows whenever no paid line is queued. It earns nothing, and nobody else ever sees it.
Technically it only adds one settings field and one background process, and sends no code anywhere. Whether your employer is fine with it is a different question, and one to ask them rather than us.