🐙
Freeadvertisingclaude code skillmarketingopen source

Claude Ads is a Claude-first skill for paid-media operations, covering Google, Meta, YouTube, LinkedIn, TikTok, Microsoft, Apple, Amazon, Reddit, Pinterest, Snapchat, and X. It runs source-grounded ad account audits with deterministic scoring, produces versioned JSON reports, and gates any account changes behind explicit capability checks. MIT-licensed, 7,900+ GitHub stars.

What is Claude Ads?

Claude Ads is an MIT licensed skill for Claude that handles paid media operations across twelve ad platforms, covering Google, Meta, YouTube, LinkedIn, TikTok, Microsoft, Apple, Amazon, Reddit, Pinterest, Snapchat and X. Its main job is the account audit: it works through an ad account against sourced criteria, scores findings deterministically so the same inputs produce the same result rather than a fresh opinion each run, and writes versioned JSON reports you can diff over time. The safety design is worth noting. Anything that would change a live account is gated behind explicit capability checks, so the default posture is read and report rather than act. That is the right default for a system with a billing relationship attached. Installation follows the normal skill workflow, meaning you add it to your agent's configuration and connect whatever platform access it needs, rather than signing up for a service.

Who is Claude Ads for?

This suits someone who manages several ad accounts and keeps redoing the same review by hand. Agencies auditing a new client's account before quoting, freelancers doing monthly checks across a roster, or an in house marketer who inherited campaigns nobody has examined in a year. The deterministic scoring is what makes it useful in that setting: a reproducible score can be tracked month over month and shown to a client, which a conversational summary cannot. The versioned JSON output matters for the same reason, since it lets you diff two audits and see exactly what moved. It fits less well if you run one small account and mostly need creative ideas rather than structural review, and it assumes you are comfortable working through an agent rather than a dashboard.

How does Claude Ads compare?

Marketing Skills is the broader library covering marketing work generally, where this is deep in one discipline and built around a repeatable audit procedure. If your gap is copy and campaign thinking rather than account hygiene, start there. Postiz is an entirely different job, organic social scheduling and publishing, which pairs with paid rather than replacing it. Dub covers the measurement end, short links with click analytics and attribution data, which is what you reach for when the question is what happened after the click rather than how the account is structured. Claude Ads is the only one of these built specifically around auditing paid accounts and gating changes to them, so choose it when review and reporting is the work.

Compare with: Marketing Skills, Postiz, Dub

Claude Ads pricing

The catalog records this as free, and the project is MIT licensed. There is no software cost and nothing to host. The costs sit around it. You are spending agent usage every time it works through an account, and an audit across twelve platforms is a substantial amount of reading. You also need whatever access each ad platform requires, which is an account and permission question rather than a fee. And the spend being audited is, of course, the real money in the picture.

Notes

No notes have been added for this tool yet.

Claude Ads FAQ

Can Claude Ads change my ad accounts?
Not by default. The project gates any account modifying action behind explicit capability checks, so the standard mode is reading and reporting. That is a deliberate design choice given that a mistake in a live account costs money immediately. Review what you are granting before you enable anything that writes.
Which ad platforms does it cover?
Twelve, spanning the major paid channels: Google, Meta, YouTube, LinkedIn, TikTok, Microsoft, Apple, Amazon, Reddit, Pinterest, Snapchat and X. Coverage that broad is unusual for an open project, though depth naturally varies by platform, so check the repository for how thoroughly the specific channel you care about is handled.
What does deterministic scoring mean here?
It means the audit applies fixed criteria rather than producing a fresh judgment each run, so the same account in the same state scores the same way twice. That is what makes month over month comparison meaningful, and it is why the versioned JSON reports are useful. You can diff two runs and see what actually changed.
Browse all Open Source