Updates documentation when the code surface drifts away from what the docs describe.
Reusable automation loops for your repo.
loopy is a library of guardrailed loops software teams import into their repositories. One contract, a fail-safe runner, and a reviewable pull request out the other end — import one in a single command.
One contract, every loop.
However smart its internals, every loop follows the same five-stage shape and runs through a single loop-agnostic runner. That uniformity is what makes loops importable, reviewable, and safe.
Loops that keep your repo healthy.
Each loop does one job and ships its own reviewable output. Add any of them with
npx loopy add <loop>.
One grouped pull request bumping your non-major dependency updates together.
Drafts a changelog entry from unreleased commits using conventional-commit history.
Posts an advisory automated review comment on a pull request — never blocking.
Backfills tests for uncovered changed lines, self-validating before it opens a PR.
Human-gated fixes for security findings above a configured severity threshold.
Self-heals docs by drafting KB articles for recurring support-ticket topics.
Z-score anomaly detection that ships a concise brief when a metric goes sideways.
Tracks overdue action items and flags recurrence across past incidents.
Import a loop in a single command.
loopy vendors the loop into your repo — workflow, config, and playbook — so it lives in your version control and runs on your CI. No platform to adopt, no lock-in.
-
+
Add a loop
npx loopy add dep-updates— vendors the loop and its guardrails into your repo. -
≡
List the catalog
npx loopy list— see every loop available to import. -
▷
Run it
npx loopy run <loop>— trigger a loop locally or wire it to CI / cron.
One framework, many loops.
Specialist tools are excellent in their niche. loopy's angle is different: a single guardrailed, importable framework that spans many loops with one contract and one CLI.
| Dependabot / Renovate | CodeRabbit | Vanta / Drata | Statsig / Eppo | loopy | |
|---|---|---|---|---|---|
| Scope | Dependency updates | PR review | Compliance | Experiments | Many loops in one framework |
| Install | Per-tool setup | Per-tool setup | Per-tool setup | Per-tool setup | One CLI — npx loopy add |
| Output | PRs | PR comments | Dashboards / reports | Dashboards / readouts | Reviewable PR or comment, uniformly |
| Guardrails | Update rules | Review config | Policy controls | Experiment gates | First-class: allowlist · caps · idempotency · human gates |
| Extensibility | Closed product | Closed product | Closed product | Closed product | ✓ Importable, open loops |
| AI provider | n/a | Vendor-managed | Vendor-managed | n/a | Provider-agnostic (defaults to OpenRouter, configurable) |
Honest take: the specialists go deeper in their niche — Renovate's update graph, CodeRabbit's review depth, Vanta's audit coverage, and Statsig's stats engine each outclass a single loopy loop on their home turf. loopy isn't trying to beat them at X; it gives you one guardrailed, importable framework spanning many loops, so you adopt the contract once instead of stitching together point tools.
Loops that remember, wait, and resume.
Beyond single-shot loops, loopy ships the substrate for stateful, multi-step work that unfolds over days — the same contract, extended across time.
Durable state
A persistent store gives loops memory across runs — they pick up where they left off.
Human-approval gates
First-class checkpoints that block a plan until a person signs off — built into the contract.
Resumable plans
Multi-step plans with wait-states and gate-blocking that survive restarts and resume cleanly.
Experiment Lifecycle Orchestrator
Adopt the contract once.
Import your first loop, review the PR it opens, and let it run on your CI. That's the whole onboarding.
npx loopy add auto-docs