Skip to content
LoopSkill
Vetted, runnable agent loops · objective pass/fail

Run a loop.
Get a verdict.

The open-source control plane for running agents for more than one client. Ship a skill into a client's environment, press run, and get an objective pass / fail — the proof no skill marketplace gives you.

Every client is a separate tenant. Their credentials never touch each other. Point a bundle at a repo you own and defects from the agents running it come back as issues there — otherwise they land in the public LoopSkill tracker.

▶ Browse the runnable registry Self-host it (open source)

Popular skills

Browse all →
FREE

buzz-mesh-linux-build

general · 13
FREE

super-memory

automation · 8
FREE

plan-for-goal

ops · 6
FREE

xitter

dev-tools · 4
FREE

brainstorming

data · 7
FREE

ascii-video

marketing · 3
FREE

faster-whisper

dev-tools · 6
FREE

startup-architect

research · 5
FREE

frontend-design

dev-tools · 5
FREE

llm-wiki-hermes

data · 2
FREE

agentic-os

automation · 2
FREE

llama-cpp

data · 1

Public bundles

Browse all →
BUNDLE

Research & Report

2 skills
BUNDLE

Corey's Marketing

49 skills
BUNDLE

LoopSkill Essentials

53 skills
BUNDLE

Dev Agent Essentials

3 skills
Install LoopSkill into your agent → copy the command
agent session
$ you
Install the loopskill skill from app.loopskill.io/skill please
$ agent
→ Fetching SKILL.md from app.loopskill.io/skill
→ Wiring LoopSkill MCP — run loops · search · install · bundles
✓ LoopSkill installed. The runnable registry + 80,000+ skills — ready.
Just tell me what you need.
$ you
Run the secret-scan loop on my repo before I push
$ agent
→ Running secret-scan-loop under enforced bounds (bounded confinement)
→ Scanned working tree · exit 0 · no leaked credentials found
✓ Loop passed in 1.2s. Safe to push. Verdict logged to the registry.
Works in Claude Code, Cursor, Cline, OpenClaw, Hermes & Windsurf — any MCP-capable agent.
Sync your fleet → Browse bundles → or compose one

Compose skills into a bundle, share one link, and every agent you run installs it. See Fleets → · Browse 55+ skills · How we audit

Runnable + verified

Every loop runs under enforced bounds and returns an objective pass/fail. Proof, not a prompt.

MCP-native

Dedicated tools. Your agent runs loops, searches, and installs from inside its own conversation.

Open source

MPL-2.0. Self-host the whole registry for free. Pro is hosted convenience, never a feature gate.

Curated + signed

Small trusted core, audited every push, ed25519-signed delivery — inside the 80k-skill universe.

Open-source meta-skill · Free skills available · Cancel anytime

80,000+
Skills across federated registries
55
Curated & signed in our catalog
pass / fail
Every loop run, objective + bounded

Production infrastructure for teams running AI in production

Marketing Reports, calendars, audits, proposals
Code Code review, refactors, test scaffolds
Web scraping Stealth scrapers, structured extraction
Ops Onboarding, invoicing, standups
Sales Outreach, CRM ops, proposals
Sim & robotics ROS workflows, simulation pipelines

Works with every
major AI agent

Claude Code
Cursor
OpenClaw
Hermes
Cline
Windsurf

How it works

Your agent already does this — it just doesn't know about us yet.

LoopSkill is a single Markdown file. Drop it into your agent's skills directory. From then on your agent can run vetted loops (objective pass/fail under enforced bounds), search 80,000+ skills across federated registries, install the right one locally, and group your picks into a shareable bundle — versioned, signed, auto-rolled-back.

01
You describe the goal Plain language, in your existing AI agent. No searches, no docs, no copy-paste.
02
LoopSkill finds the skill The meta-skill makes your agent aware of every LoopSkill skill. It picks the perfect match.
03
Installs locally in seconds Skill files written to your machine. Your credentials, your data, your control.
04
Results delivered A report. A scrape. A calendar. A patch. Done — and ready to ship.
~/clients/X — agent session
$ you
install the loopskill skill from app.loopskill.io
$ agent
→ Fetching SKILL.md · wiring MCP (run loops · search · install · bundles)
✓ LoopSkill installed. Runnable registry + 80,000+ skills — ready.
$ you
build my agents a memory + scraping stack and keep them all in sync
$ agent
→ Searching ClawHub · skills.sh · curated — match: super-memory + agent-browser
→ Composing bundle "agent-core" (2 skills, 1 link)
→ Subscribing fleet on stable channel — 3 agents reconciled
→ Auto-rollback armed; only changed files pulled (304-cheap)
✓ Fleet in lockstep. Every machine, same signed stack.
elapsed: 71s

For creators & operators

Publish a loop. The whole world can run it.

LoopSkill is open source (MPL-2.0). Publish a vetted, runnable loop to the registry and anyone can run it under enforced bounds and rate it — earn 50% recurring rev share on every Pro referral.

Learn about the creator program

Two free steps to value

Two free skills.
That's the whole onboarding.

No card. No subscription. Each step does something useful on its own.

  1. FREE FOREVER step 01 — meta-skill

    Install loopskill-skill

    One phrase to your agent. It gains registry awareness — it can run loops, search, and install any skill in the library.

    "Install the loopskill skill from app.loopskill.io/skill" Install instructions
  2. FREE · MIT step 02 — first skill (open-source gateway)

    Add super-memory

    Cognee + Postgres + pgvector wired together. Your agent gets durable cross-session memory in 60 seconds. MIT-licensed — yours forever, no subscription. The pro_open gateway: install free, see how we ship, then upgrade for the rest of the catalog.

    "Add the super-memory skill" View skill