The Loop
From chat to merged code in under 12 hours
One message in Slack starts a pipeline that writes the fix, opens the MR, reviews it, routes it to approval or auto-merge, and posts the morning scorecard — all on your compute, on your meter.
Six steps
How the loop runs
Ticket
Type /ticket in Slack or apply a status::now label to any issue. The requester is done — everything after this is automated.
Dispatch to your CodeBuild
The loop picks up the ticket and spins a fresh runner in your AWS CodeBuild. Your compute, your logs, your network — nothing leaves your account.
MR with manifest
The agent opens a merge request with a complete manifest: files touched, tests run, risk tier, and a plain-English summary an approver can act on without reading every line of diff.
AI review verdict
A second agent — still in your CodeBuild — reviews the MR independently and labels each change: auto-merge candidate, needs-human, or blocked.
Auto-merge lane or approver card
Auto-merge candidates close on the next pipeline run. Everything else posts a concise approver card to Slack — one-click approve, hold, or reject. No tracker, no inbox.
Morning scorecard
At 07:00 every morning the loop posts a scorecard: tickets closed, MRs merged, blocked items, and infrastructure cost delta. One message, full context.
Runs on your meter, not ours
The loop runs entirely inside your AWS account. Infralista never proxies your source code, credentials, or agent logs.
CodeBuild: your account
Agent runners spin up in your CodeBuild, billed at standard AWS on-demand pricing. No Infralista compute markup, no persistent servers sitting idle.
Claude: your subscription
Agents run on your Claude subscription (flat-rate) or through your own Amazon Bedrock (per-token). Your bill, nothing proxied through Infralista.
Approver seats, unlimited requesters
Pricing is per account plus the number of approver seats — the people who can merge. Requesters who open tickets are unlimited at no extra cost.
What Infralista never sees
Your source code, AWS credentials, agent logs, approval decisions, or any data processed in the loop. The platform configuration is the only thing we ship.
Live throughput from our own app
Chócala — a multi-tenant coaching platform with real daily users — has been running the loop since August. These are the actual numbers from the production agent sweeps.
| Period | Tickets in | Merged | Merge rate | Points |
|---|---|---|---|---|
| August | 430 | 174 | 40% | 1,096 |
| September | 288 | 138 | 48% | 783 |
| Oct 1–17 | 569 | 443 | 78% | 1,954 |
| Total | 1,287 | 755 | 59% | 3,833 |
* Data from the Chócala production app, August–October 17 2026. "Points" are weighted story points scored at agent-sweep time. Merge rate counts MRs fully merged to main, not just approved. The merge rate improvement from 40% → 78% reflects loop tuning over the three-month period. A live board replaces this static table in week 4 of the 30-day build.
Ready to close tickets while you sleep?
The loop runs in your AWS account, on your meter, with your approval gates. Book a call to see it running on our own app.