* CLI and AI-friendly. this toolkit is being run by our AI agents daily, they use it to research, then update CSV files, log their decisions in git log and then we review PRs and apply updates to campaigns
* export/import data in CSV, store full config of campaigns, keywords, creatives in JSON/CSV
* does not require API access (which is hard/impossible to get for certain Apple Ads Orgs)
this allows us (just me really, I run all the marketing ops...), to scale of what otherwise would be impossible. and allows to fight instability/randomness of Apple Ads performance (git log contains full history of decisions, context, reasons, hypothesis).
inspired by Go analysis framework, I developed multiple "linters" that detect problems with setup, ensures best practices. you run apple-ads analyse keywords and get colorful rich ASCII with histograms, CPI/CVR. I am using very basic statistics for now (nothing beyond simple Bayesian was needed for now, which is why it is in Go).
the power of it, is that normally AI agents (say Claude Opus 4.6) would repeatedly write ad-hoc jq, awk, grep or low-quality Python scripts, which kind of the same. Now, I assembled all these learnings and what AI typically does, cleaned it up, made it better to use (based on typical problems when running it with AI agents in loop), things like time filtering, no-color output, help and documentation (inspired by Go analysis framework I made it self-discovery) so that your agent can just run apple-ads -h and learn how to run it instantly.
It is very much "Ads GitOps". It is is really effective, works well... and free. I think should belong to community. so here you go! if it helps someone save couple bucks that would made my day!
https://github.com/ndx-technologies/go-apple-ads