The open registry of runnable agent artifacts
Spotify for AI agents. Browse, pull, and deploy skills, bundles, loops, and personalities to your fleet. Loops run on schedule — autonomous, verified, reporting outcomes while you sleep. Claude, Codex, Cursor, any MCP client. Clone it, value in 60 seconds. No signup.
Paste that install line into any agent chat, or browse app.loopskill.io directly. No credit card, no account, nothing phones home.
One library, four artifact types
Everything your agents pull from, in one place
Skills and bundles are config. Loops and personalities are runnable — the clone-to-wow artifacts no other registry has.
Skill v1
A single capability — SKILL.md plus scripts and references. Install once, your agent gains a new power.
Bundle v1
A curated set of skills you deploy and sync to a whole fleet at once. Version it, share it, roll it out.
Personality v1
A deployable persona — system prompt plus config. Give an agent a voice, a discipline, a way of working.
Loop v1
A safety-bounded agent loop — success condition, verification, stopping criteria, budget. Schedule it on your fleet and it runs autonomously on cadence, reporting pass/fail outcomes via telemetry. Or test once with a hosted run.
The white space
A vetted registry of safety-bounded loops
People hand-roll agentic loops and they spin out — runaway processes, $400 overnight bills. No vetted loop library exists. LoopSkill loops carry their safety envelope before you pull them: max turns, tool allow-list, a verification script, and explicit stopping criteria. And the registry doesn't just store that contract — it runs the loop's verification and hands you an objective pass/fail, inside the bounds. Pull a loop, run its check, trust the result.
Self-host, the star engine
From git clone to a running registry in 60 seconds
The whole product is open and runnable from a clone. The paid plan is convenience, never a feature gate.
Clone & run
docker compose up — or the curl one-liner. Zero secrets, zero signup. A full registry boots on SQLite.
Deploy a loop to your fleet
Loops are autonomous operating artifacts. Schedule one and your agents run it on cadence — fixing repos, triaging PRs, watching systems — reporting objective pass / fail outcomes via telemetry. Or test once with POST /api/loops/{slug}/run to verify the bounds first.
Browse & install
A starter catalog of skills, bundles, loops, and personalities is seeded on first boot. Install one to your local agent over MCP.
Deploy to your fleet
Push a bundle to every agent you run and keep them in sync. Self-host forever, or let us run it for you.
Simple, honest pricing
Free to self-host. Pay only for convenience.
Every feature is in the open-source repo. Pro is just “don’t run it yourself.”
- ✓ The entire product, open source (MPL-2.0)
- ✓ Skills, bundles, loops, personalities — all of it
- ✓ Deploy & sync to your own fleet
- ✓ Your data, your infra, nothing phones home
- ✓ We run it — managed, updated, backed up
- ✓ Managed sync across your fleet
- ✓ Private registry for your own artifacts
- ✓ Zero infra to maintain
Ship features while you sleep