Loopsmith
Loopsmith is "the Chief of Staff that builds your Chief of Staff." Every executive wants an AI agent for their workflow but can't architect one.
Core trick
A Claude-graded quality gate scores each loop output and writes lessons to memory that feed the next run's prompt.
Description
Loopsmith is "the Chief of Staff that builds your Chief of Staff." Every executive wants an AI agent for their workflow but can't architect one.
Loopsmith interviews a non-technical leader in plain language, then designs and runs a self-improving "operating loop" — and hands them a working
Claude Code repo they keep.
The loop ingests real signals (Slack / email / calls / calendar), decides what matters, drafts the actions, grades itself against a quality gate that
holds weak output, and writes durable learnings so the next run is better. The self-improvement is measurable, not cosmetic: on a fixed input the gate
score climbs run-over-run (75 → 90 → 97 → 98) as the loop learns the operator's preferences, and the run history proves it on screen. You never write a
prompt, pick a model, or hear the word "sensor" — you just talk. Built entirely during Claude Build Day; powered by Opus 4.8.
README excerpt
# Loopsmith
> **The Chief of Staff that builds your Chief of Staff.**
Loopsmith interviews a non-technical leader about a workflow they want automated, designs the
self-improving **operating loop** they need, runs it live, and hands them a working Claude Code
project they keep — one that ingests real signals, decides what matters, takes action, checks
itself against a quality gate, and writes back what it learned so the next run is better.
You never write a prompt, pick a model, or hear the word "sensor." You just talk.
Built at the **Claude Build Day** (San Francisco, June 13 2026). Powered by **Claude Opus 4.8**.
---
## Try it in 60 seconds
1. Open **`/build`** and describe a workflow in plain words
(*"Turn my Slack, calls and email into the 2–3 moves that move revenue, and draft the outreach."*).
2. Answer ≤5 quick questions — click the suggested chips, no typing required.
3. Watch your loop **run live**: it ingests signals → decides → drafts actions → grades itself → learns.
4. **Run it on your real data** — paste your actual week (Slack snippets, emails, call notes); every
move is grounded in *your* content. Nothing is sent; actions are drafted for your review.
5. **Make it yours** — tell it what you'd change ("always lead with the dollar amount"); it applies
that immediately *and* remembers it forever. Watch the gate score climb run over run.
6. **Download the project** (`.zip`) — a runnable Claude Code repo, or let the scheduler run it every
Monday a