Trigger your Software Factory automatically from events that matter

Loops watch tickets, alerts, CVEs and schedules, screen out work the factory shouldn’t take, and keep the rest moving overnight and on weekends. Configure each Loop once and reuse it across teams.

Abstract illuminated glass loop with incoming event streams and an approval gate.

Trusted by teams operating AI in production

Capgemini logo
Vattenfall logo
hear logo
bunq logo
Moneybird logo
Yoco logo
Moss logo

Run the right work around the clock

WATCH

Start work from the systems your teams already use

A Loop starts from a ticket in your issue tracker, a monitoring alert, a security advisory (CVE), a schedule or a webhook, and one Loop can combine several triggers. It connects to your existing repositories, issue trackers and CI/CD with the same policies and permissions, so nobody has to pick the work up by hand.

LoopsTemplates

Templates

  • Triage bug reports

    SlackGitHubWatches a Slack channel for bug reports, finds the root cause and opens a fix PR.

  • Investigate escalations

    LinearSlackWhen support escalates to engineering, Factory finds the root cause and suggests a fix.

  • OWASP security hardening

    GitHubWeekly scan for OWASP Top 10 issues: injection, XSS and insecure defaults.

  • Investigate alerts

    SlackWebhookWhen an alert lands in your channel, Factory pulls metrics, logs and traces.

  • Fix CI failures

    GitHubReads the build logs of failing checks, finds the cause and pushes a fix.

New task
ScreeningCan Factory take it? Should Factory take it?
  • Scope
  • Risk
  • Guardrails
Pass to factoryNext steps
  • Specification
  • Planning
  • Coding
  • Review
  • Testing
  • Merge
Screened outNot suited to the factory
New task
ScreeningCan Factory take it? Should Factory take it?
  • Scope
  • Risk
  • Guardrails
Pass to factoryNext steps
  • Specification
  • Planning
  • Coding
  • Review
  • Testing
  • Merge
Screened outNot suited to the factory

TRIAGE

Screen every task before the factory spends on it

Before any work starts, the Loop checks whether the factory can and should take the task. Work it wasn’t built for gets screened out, so you don’t pay for low-quality pull requests. Tasks that pass move into specification, planning, coding, review and QA, security testing and merge, and costs are counted per accepted change.

RUN

Make progress on routine work while your team is offline

Bug fixes, dependency upgrades and CVE patches keep progressing outside working hours. For each change type, your team decides whether agents act alone, a human reviews, a human approves, or the work stays human-owned. Work that needs a sign-off waits at that gate, ready when your team is back.

LoopsNightly dependency upgrades

Runs by hour
Runs128+18%
Success rate94.53%+2.1pt

LoopsNightly dependency upgrades

Runs by hour
Runs128+18%
Success rate94.53%+2.1pt
Central LoopOne shared definition
  • Triggers
  • Guardrails
  • Approval gates
  • Skills
Team A
Team B
Team C
Central LoopOne shared definition
  • Triggers
  • Guardrails
  • Approval gates
  • Skills
Team A
Team B
Team C

REUSE

Set up a Loop once and share it with every team

A Loop’s triggers, guardrails, approval gates and skills live in one central place. Other teams reuse it instead of rebuilding it, which keeps agent behavior consistent and stops agent sprawl. Review verdicts and failures feed back into the factory’s skills, so each Loop gets better with use.

FAQ

Frequently asked questions

Does a Loop run the pipeline steps in a fixed order?

No. A factory mirrors your team’s own delivery steps, whether that’s three steps or eight. You describe the workflow in plain language and point each step to the skills, MCP servers and subagents it should use.

Do Loops come with ready-made setups, or do we configure everything ourselves?

You write each Loop’s instructions and choose its triggers, model, reasoning effort, skills and MCP servers.

Can a failed CI pipeline or end-to-end test start a Loop?

Yes. A failed CI run can start a new session, so an agent picks up the failure and works on a fix. Any other system can start a Loop through a webhook.

Can Loops connect to our own build and deploy tools?

Loops connect to your existing repositories, issue trackers and CI/CD. A tool without a built-in integration, such as an on-premises build server, can start a Loop through a webhook and pass in the event.

Who can see what a Loop is doing?

Every Loop run opens a shared session that all members of that factory can follow and comment on. Sessions a person starts on their own stay private to that person.

Do Loops replace the tools we already use?

No. Loops listen for events in your current tools and send the finished work back to them, usually as a pull request in GitHub or GitLab.

What happens when one event contains several problems?

After triage, the Loop can open a separate session for each problem. Each session then moves through the pipeline on its own.

Set the direction and let the factory deliver

Loops are where the work starts. AI agents then plan, build, test, review and monitor it, and you approve what matters.