# 1,300 emails in 10 days: How Matt Abate's Claude agents did builder outreach for his open-source repo

Matt Abate ran his personal assistant as a swarm of Claude agents with job titles, and two of them did outreach for his open-source repo. A CTO agent read other builders' repos and left GitHub issues. A CMO agent emailed each builder whose repo had been read. He says he sent 1,300 emails in 10 days, 320 builders wrote back, and the repo went from 21 stars to 348. On day 10 the agents sent over 300 emails from his personal Gmail, and Google disabled the account. He got it back the same night.

**Try this step:** <strong>Give the assistant job titles.</strong> In his words: "My assistant is a swarm of agents with job titles." Two of them handled this project: a CTO and a CMO.


| | |
|---|---|
| **Who** | Matt Abate, New York; his X profile lists TextQL. X @_laundry_matt_ and GitHub mattabate (repo mattabate/life-distributable), both from his own thread, verified 2026-11-18 |
| **Problem** | Telling builders about a new open-source personal-agent repo |
| **Stack** | Claude Code agents with job titles (CTO, CMO), GitHub issues, a CRM of builders, email from his personal Gmail |
| **Confidence** | Medium (his own public thread and repo; self-reported; no outside check of the counts) |
| **Last verified** | 2026-11-18 |

## The problem
His open-source repo is a self-hosted team of AI agents built on Claude Code. He does not state his goal in the thread. The plan he describes is for the CTO agent to read builders' repos and learn from them, and for the CMO agent to leave issues and email the builders. He then measures the result in stars.

## The blueprint (as he describes it)
1. **Give the assistant job titles.** In his words: "My assistant is a swarm of agents with job titles." Two of them handled this project: a CTO and a CMO.
2. **Have the CTO agent read other people's repos.** It read the repos of 900 builders and pulled the good ideas into his own.
3. **Report back through GitHub issues.** The CTO left 1,200 GitHub issues describing what they had fixed on their side.
4. **Have the CMO agent keep a CRM.** It kept a CRM of 3,600 builders and emailed each one whose repo had been read.
5. **Track who replied and who starred.** His chart splits 1185 builders emailed between Oct 28 and Nov 6 into email repliers, issue-only repliers and no reply yet, and counts the stars from each group.
6. **Turn replies into an event.** He and a co-host set up a builder meetup in New York on Fri Nov 6 (his event page says "Estimated 20 people"). He got the Google text there.

## Stated results
- **Email volume:** "In 10 days I sent 1,300 emails."
- **Replies:** 320 builders wrote back (27% of the 1185 emailed). A further 160 (14%) replied on an issue only. 705 (59%) had not replied yet.
- **Stars:** the repo went from 21 stars to 348. By his chart, 217 stars came from the email repliers, 34 from the issue-only group, 49 from builders who had not replied, and 48 from people the agents never emailed.
- **Event:** a builder meetup in New York, which he says came out of the outreach.
- **What broke:** "On day 10 we sent over 300 emails from my personal Gmail. I got banned, but I got the account back the same night."
- His own takeaway: "I don't know if there is a lesson here: but 300 emails a day from your own Gmail is probably too many."

## Caveats
- This is his own public thread and his own repo. Nobody else has checked the counts.
- The post says "1,300 emails" and "300 stars"; his chart says 1185 builders emailed and 348 stars. The thread does not reconcile 1,300 emails with 1185 builders, and the first post's "300 stars" with the chart's 348. This page uses the chart and later post figures for stars.
- Stars are a vanity metric. They show interest, not installs or use. His chart shows the stars by reply group, but it does not show whether the emails caused them. 48 stars came from builders who were not emailed.
- The repo's README asks readers to star it to get weekly addenda, so stars were also a way to receive updates.
- The thread says the Gmail account was disabled and restored the same night. It does not say what Google flagged or how the account came back.
- The thread does not say whether a person approved each email before it went out. His README says actions that contact a person need an approval step, but the thread does not say how outreach was handled.
- The thread does not describe the agents' prompts, how the CRM was built, or what the emails said. It does not say what the 1,200 GitHub issues contained.
- The repo is free and MIT licensed. The README says costs go to outside services (Claude, Apple, Tailscale, Backblaze) and "Nothing goes to me." The agents run on Claude Code, which needs a Claude plan.
- Handles: X @_laundry_matt_ and GitHub mattabate (the repo owner, linked from his thread) are verified from his own posts. His website is mattabate.com (linked from his X profile). No other platform was checked.

## FAQ
### Which AI tool did Matt Abate's outreach agents use?
Claude. He calls them "my Claude agents," and his repo describes a swarm of agents built on Claude Code.

### What went wrong with his Gmail?
On day 10 the agents sent over 300 emails from his personal Gmail and Google disabled the account. He got it back the same night. His own advice is that 300 emails a day from your own Gmail is probably too many.

### How many builders wrote back?
320 replied by email, 27% of the 1185 builders emailed between Oct 28 and Nov 6. Another 160 replied on an issue only.

### Did the outreach cause the stars?
His thread does not say. His chart shows 217 of the 348 stars came from email repliers and 48 came from builders who were never emailed.

## Sources
- https://x.com/_laundry_matt_/status/2119100902380335537 (primary: his own four-post thread, Nov 7, 2026, with his chart, event page and Google text screenshots)
- https://github.com/mattabate/life-distributable (primary: his repo README and licence, checked 2026-11-18)

---
*Want agents like these working your distribution? [Get It Built](/get-it-built). Get new teardowns weekly: [subscribe](/subscribe) (double opt-in).*