The Apollo MCP will run your entire outbound motion. It just has no idea how.

Point it at your account and it will run any search, build any sequence, and send any email you ask for. It will also let you target the wrong people, write copy that reads like a robot, and burn your sending domain. It does what you say, not what you meant. The judgment to run outbound well was never in the tool. It was always in the operator.

So we wrote the operator down.

What Apollo Operator is

Apollo Operator is a system of 13 open-source skills for Claude Code that turn the raw Apollo MCP into a guided outbound motion. It is free, it is public, and it runs on your own Apollo account.

One router skill loads first, reads where you are in the process, and hands you to the right specialist. The other 12 cover the full motion by stage:

• Targeting: turns a written ICP into a real Apollo search and a scored targeting profile, so "mid-market SaaS CROs" becomes filters that return the people you actually meant.

• List building: searches, enriches, and verifies, then grades the list against your ICP before anything sends. A list that looks clean can still be full of the wrong people.

• Copy and sequences: builds short, human sequences matched to the persona, then reviews every step for spam risk before it goes anywhere near a send button.

• Multichannel (optional): layers LinkedIn and calls on top of email when the motion calls for it, instead of pretending email is the only channel.

• Go-live: enrolls the list, requires a human approval, and pulls contacts out of the sequence the moment they reply.

• Deliverability: enforces warmup, verification, and the bounce rules that keep you in the inbox.

• Iterate and signals: scores replies for real intent instead of raw volume, structures clean experiments, holds a weekly cadence, and finds the reason to reach out now.

Every skill holds your campaigns to the same standard we hold ours. The judgment we apply on client accounts, who to target, how to grade a list, and when a send is a risk, is written down and made runnable.

Built by operators, tested live

We did not write this from theory. We built it on our own Apollo account and ran it, and the testing paid for itself immediately.

Three examples. A single title filter quietly inflated one search from 679 real matches to 91,649, the kind of mistake that fills a campaign with the wrong people and looks like progress while it does it. A bulk create call returned a cheerful success message while silently creating contacts we could never actually use. And a 2,591-row list that passed every mechanical check came back 33% off-ICP the moment the scorecard graded it against the actual targeting profile. Clean data and right data are not the same thing.

We caught all of it. The fixes and the checks are encoded in the library, so you never have to catch them yourself. The tool will not tell you when you are about to make an expensive mistake. The operator will.

Deliverability is the one place we do not bend

Most of Apollo Operator surfaces a risk and then lets you decide. Deliverability is the exception. Warmup, verification, and a hard bounce ceiling are not suggestions, because you cannot out-write a reputation problem. Once your domain is burned, the best copy in the world lands in spam. The library gates on this so you do not learn it the hard way.

Who it is for

Founders and go-to-market teams running outbound on Apollo who want a system, not a black box. Open the folder in Claude Code, describe what you want, and the router takes you from there. Use it free to run your own motion.

If you would rather have it built and run for you, that is what we do at Creatop. Either way the method is the same, and now it is public.

Get Apollo Operator on GitHub. Use it, break it, and tell us what you find.