How Claude Code Ads Converts Your AI “Thinking” Time into Actual Earnings
The most profitable piece of real estate in development is not the code editor or even the surrounding user interface. It is the single line of text that appears while Claude “thinks.”
That specific slot is what the new Claude Code Ads product focuses on.
The platform replaces the standard, throwaway spinner text that shows while your AI assistant processes your commands with clear, concise messages. Developers who use Claude Code earn 75% of all ad revenue that appears on their terminal.
If you were ever looking for a way to generate income while coding without the burden of starting a secondary business, this is exactly how it appears.
Defining Claude Code Ads
Claude Code Ads operates as a small-scale ad network that is built directly into the Claude Code CLI and editor plugins (such as Cursor or VS Code). When Claude initiates a command, be it reading an existing codebase, refactoring, or debugging, it swaps the standard “Discombobulating…” spinner with a sponsored prompt, like:
✶🧠 Acme DB, Postgres that scales itself. (esc to interrupt)
Following that, the second line explicitly displays that the message is an advertisement and indicates the specific earnings from that moment:
ad·🧠 Acme DB, Postgres that scales itself ↗ +$0.0038
It is precisely that simple. There are no banners, no pop-ups, and no interruptions to your terminal experience; the system only displays ads on the single line that you already view whenever Claude is calculating something.
Product-wise, Claude Code Ads sits at the intersection of monetizing developer tools and running a dedicated Claude terminal advertising network. It provides:
A dedicated ad platform for companies that wish to market their products to software developers while those developers are actively working on tasks.
A mechanism where individual software developers or development teams can earn money by having ads show up on their terminal while they work.
Mechanics of replacing the terminal “thinking” line
The fundamental concept is extremely simple, yet impactful: the system takes over your terminal “thinking” lines with ads in a seamless manner that does not alter your codebase or workflow.
Essentially, the Claude Code Ads extension:
Integrates with Claude Code’s spinner and status line This extension leverages the configuration hooks that are documented by the product, like the
statusLineandspinnerVerbssettings located within~/.claude/settings.json. There are no obscure hacks involved; Claude Code remains unchanged from before, other than the fact that while you wait on an active session, the spinner text is replaced with a sponsored line.Makes an API call to Claude Code Ads auctions during wait states When a session enters a “thinking” stage that is long enough to matter, the extension makes a small call to the Claude Code Ads server: “I have a verified human-initiated impression on this machine; what ad won?” The server holds an actual auction between all the active campaigns that line up with your location and any other targeting filters. The top effective bid gets that impression.
The winner gets rendered as one short ad string in the terminal The winning ad shows up as a compact line of text (3–60 characters) plus the ad’s destination URL. Your terminal status line or Claude Code spinner verb displays it (explicitly labeled as an ad). On terminals and editors where it’s supported, that line is a clickable link.
We count verified views, not garbage traffic To ensure a trustworthy marketplace, impressions are only tallied when:
A person ran a real Claude Code command.
The ad stayed visible for at least some minimum duration (usually about ten seconds, depending on settings).
The user session is active, not just an idle window in the background.
This “verified views” approach is why serious advertisers find the platform usable. Automated traffic and incidental terminal blinks get blocked out; only real, in-progress execution windows count.
From the developer’s side, there’s no change in how you write code. You just fire up Claude, wait for the thinking phase, and now you get paid in your status line.
75% of ad spend is paid to developers
Most “earn while browsing” projects fail because the user sees pennies while the middleman pockets nearly every cent. Claude Code Ads does the opposite.
The revenue model is simple to the extreme:
Advertisers pay Claude Code Ads for impressions (clear CPM pricing, sold in blocks of 1,000 verified ten-second views).
Exactly 75% of that ad spend goes straight to the developers who served those impressions.
The other 25% goes to cover the platform’s operational costs: running the marketplace, moderation, billing, and payouts.
So, if a campaign bids $4.00 CPM, $3.00 of that flows into the developer pool. Your terminal gets a share of that pool proportional to how many qualified impressions it delivered.
There are no obscure point systems, no convoluted conversion math, no hidden rounding tricks. Your dashboard displays:
Impressions delivered
Effective CPM and your net CPM after the 75% payout
Earnings by session and by day
Since price points are advertised and billed in explicit CPMs, advertisers can forecast cost-per-developer-session easily. Similarly, developers can make an estimated prediction of a monthly yield resulting from their Claude Code sessions. If you run heavy use, say dozens of longer Claude tasks each day, it could result in several thousands of impressions per month; assuming an average mix of bids, this results in a relatively modest but reliable reduction in the costs of AI subscriptions and cloud-based services.
Incentives for user clicks
Even though Claude Code Ads operates principally on impressions, clicks are relevant as well.
This is because clicks are the strongest indication that the ad has caught the attention of a developer within the flow of their work. Claude Code Ads takes this metric and utilizes it in two distinct ways:
To advertisers: performance-driven optimization The platform is structured in a way such that ads that receive higher CTRs with actual developer traffic have a higher probability of being selected as the ad of choice, relative to other ads at the same bid price because the ad is delivering more value to the advertiser. The implication is that the advertiser will want to present relevant offers to developers; links to documentation, tutorials, and sandboxes or playgrounds; developer tiers that are free to try; debugging software, database management systems, and similar products and services.
To developers: a click-based incentive system Impressions form the baseline payment structure, but in addition, the model for clicks that exist on this network is very simple, transparent, and additive:
You will earn the same amount that your advertisers do per click, whether it was billed at a CPM (for example, $10 per CPM) or a CPC (for example, $0.10 per click) on their end.
You will receive 75% of that amount per event, meaning that if that click is worth $0.40 on the advertiser end, you will earn $0.30 for that click.
In the event that you click on a brand-new campaign on one of your machines, you’ll also be paid a one-time bonus for that click. This should incentivize you to click, but because it’s only a bonus, the user is never encouraged to farm the network by clicking on ads they have no interest in.
The important aspect here is that users never feel they’re being asked or pushed to click on ads. The line is short, discreet, and labeled; you’ll click on it only when there’s a resource or tool that genuinely seems to be of interest to you. This keeps the advertisers happy, because it means the ads they are paying for will be seen by people who have some interest in their products and services. It keeps your terminal looking like a terminal, and not a casino.
The advertising model works for advertisers because
The usual approach to advertising to software developers consists of three rules:
Developers will ignore standard ad units. Most will use ad blockers or some other mechanism to prevent ads from appearing. Many developers will consider ads to be background noise in the software ecosystem.
Developers don’t spend all day scrolling. Developers spend a large amount of time in their integrated development environments, terminal windows, documentation pages, or package management systems—most developers aren’t spending a significant amount of time in the feeds of social media sites.
Developers value practicality. A simple link to useful software, documentation, or tutorials will always attract significantly more attention from developers than a sensationalized clickbait headline, for instance.
This is all very applicable to Claude Code Ads, because the model is tailored to accommodate all of those constraints. This is first about workflow. Claude Code sessions are where developers are already playing with frameworks, libraries, infra, and AI-automation, and so the appearance of an ad for Acme DB, which is "Postgres that scales itself," in the very micro-moment they're awaiting a migration script is about as contextual as you'll get.
Second, there's no bot traffic, or very little of it. Impressions can only occur when someone actually runs a Claude Code command from within a TTY or editor, so fake traffic is near-zero. You aren't paying for "developer eyeballs," but for genuine developer attention within genuine dev sessions.
Third, ads are kept short. There's very little room for brand puffery in 3–60 character lines. What wins is ads with one-liners with a sharp value proposition: a better error tracking tool, an auth service built with DX in mind, a simple metrics stack.
All of this is a differentiator on the best of developer ad platforms, from daily.dev Ads to Idlen's in-IDE network to EthicalAds. Where Claude Code Ads reaches the subset of devs that's already buying into AI code agents and their workflow. That group is heavily skewed toward founders, senior engineers, and early adopters who are building the next generation of tooling.
How to earn doing terminal ads in practice
Assuming your goal is to get as easy as possible, then the workflow is simple:
Sign up for free You install the Claude Code Ads extension for VS Code, Cursor, or VSCodium, or run the Claude CLI from the command line. The first time you use it, you have to log in via the browser and set up a payout method. Otherwise, you will see a "sign in to start earning" link in the spinner instead of ads, so you won't be monetizing your terminal sessions inadvertently.
Go about your normal work Every time you launch a Claude Code command, which will block for some time, Claude Code Ads will scan for a winning ad. If there's an ad for you, it will take the place of the terminal prompt while thinking for the length of the blocking event. In the background, each verified view will be credited to your ad revenue at the prevailing market rate.
Collect your revenue Revenue can be seen in a little status bar, as well as a larger dashboard which shows your estimated revenue for today, month-to-date, and all time. When your balance goes above the minimum payout threshold, you're eligible to withdraw from your preferred method.
Turn off ads There's also no obligation. If there's a machine or session that you'd rather not show ads in, you can disable the extension, or just use the command palette to disable it. Claude Code will immediately snap back to its default spinning wheel behavior.
Unlike many so-called "make money from your AI" platforms, Claude Code Ads was purposely created to be boring: there are no tokens or multi-level marketing programs, it is CPM-based, with a very high, 75% developer cut.
Claude Code Ads vs traditional developer ads
When looking at Claude Code Ads vs traditional developer ads, the differences come back to surface, economics, and trust.
Surface The traditional ad channels, such as LinkedIn, standard display and search are still important, but they have always been outside of the build. This means a significantly lower number of impressions are required for value. A Claude terminal advertisement, however, happens directly within the build of the agent.
Economics Typical ad channels take 50 to 80 percent of total ad spend. Here, that ratio is flipped, with 75 percent of all ad spend going to developers. This is a very different value proposition for anyone considering a Claude CLI monetization strategy or how to monetize developer apps without selling user data.
Trust On developer-native platforms, trust is the currency. Claude Code Ads won't read your prompts or code, it won't send telemetry on your code or applications to the cloud, it clearly labels all of its messages as ads, and its value comes from not taking over the developer experience, by slipping ads into developer interfaces.
A pragmatic way to monetize developer tools
If you've spent time building a tool on Claude, if you are maintaining a large open source project, or even if you just sit inside the CLI for hours a week, it can be a very good answer to "how to make money by coding," without launching a SaaS or having to hunt for sponsors.
For advertisers, it offers a unique channel for advertising to software engineers when they are at the peak of their focus on the problem, with simple pricing and minimal wasted impressions.
For developers, it is one of the purest "earn money with CLI ads" programs currently available; it doesn't change your behavior, it doesn't have confusing economics, it handles your privacy well, and your revenue scales with the amount of time you already spend inside Claude Code.
As we approach the day when AI coding agents become standard for any serious project, this ad real estate—the spinning line, the terminal indicator bar, those moments of downtime between commands in the CLI—will likely quietly become the most sought-after inventory in software advertising. This is simply Claude Code Ads' attempt to be first to market with that vision, and be brave enough to give most of the revenue to the people whose attention creates the demand.