Every loop. Every decision. Every time.
AI systems that close the loop.
Looptail watches every agent decision, scores it against your standards, ships the fix, and keeps a signed record of every pass. Observe. Understand. Improve. Automatically.
A trail in 10 minutes.
One SDK call. Every decision your agent makes becomes a signed loop event.
01 · install
pip install looptail02 · wrap your agent
import looptail
looptail.init(app="support-agent")
@looptail.trail
def handle_ticket(ticket):
...03 · open your trail
npx @looptail/cli verify \
--app support-agent --since 24h
✔ 1,204 loop events
✔ chain intact · signatures validIn private beta with design partners in regulated industries. Names when they say yes.
Every loop leaves a tail.
The Tail is an append-only, cryptographically signed record of every loop: what was observed, how it was judged, what changed, who approved it, and what happened next.
When someone asks why the system did what it did on March 3 — and what you changed afterward — you answer with evidence, not a search through three tools and a Slack thread.
- 14:03:12 observe support-agent · ticket #4211 answered · outcome pending sig 9f3a…c2 ✓
- 14:03:14 evaluate rubric refund-policy-v3 scored 0.94 · pass sig 41bc…88 ✓
- 14:07:02 issue cluster #18 "partial refunds mis-quoted" +1 case sig 77e0…4d ✓
- 14:21:40 improve patch #91 proposed · prompt diff +12 −4 · replay 132/132 pass sig b8d1…09 ✓
- 14:29:05 approve patch #91 approved → canary 5% sig e42f…7a ✓
Start free. Stay honest.
Tail Lite is free: 10k loop events a month, 30-day retention, export anytime. Paid plans add continuous evals, replay, and signed long-retention trails.