LugemLugem
The AI-asset registry, delivered as SaaS

The AI-asset registry for your enterprise

Publish once, govern centrally, consume everywhere. npm and Docker Hub for your AI building blocks — authored, reviewed, versioned, and trusted in one governed place, so every harness your team uses gets the same audited, signature-verified pull.

pdf-extractor

Skill

v1.2.0

Extracts structured text and tables from PDF files.

Verified by platform teamGroupGitHub
Signature verified · audited pull · from your storage
The same governed pull, everywhereClaude CodeCursorCopilotn8nCI pipelinesInternal harnesses
Why lugem

One governed home for every AI building block

Skills, agents, MCPs, workflows, prompts, custom instructions, and policies — reviewed, versioned, and trusted here, then distributed everywhere. Three commitments make it enterprise-ready.

Your content, your infrastructure

Lugem stores metadata and a content reference — never your bytes. Point it at GitHub, S3-compatible storage, an HTTPS endpoint, or AWS Agent Registry through one pluggable adapter, and swap backends later without changing how governance works.

The right people see the right assets

Every asset — skill, agent, MCP, workflow, prompt, or policy — is visible org-wide, scoped to a group, or kept private, enforced uniformly on browse, detail, and every pull.

Governance built in, not bolted on

One publish gate — metadata validation, secret scan, admin approval — for skills, agents, MCPs, workflows, prompts, and policies alike, over an append-only, per-org hash-chained audit trail. Every pull is an event.

Seven asset types

Everything your team builds with AI

Each type is registered like a package — with a version, an owner, dependencies, and a content reference — and consumed by whichever harness understands it.

Skills

Bundled files and scripts, installed like a package.

Agents

Agent definitions any runner can execute.

MCPs

Server manifests for any MCP client.

Workflows

Composed, multi-step recipes.

Prompts

Reusable, versioned templates.

Custom instructions

House rules and context for IDEs and agents.

Policies

Guardrail definitions enforced in CI.

How it works

From publish to pull in three steps

lugem is the source of truth and the distribution point. Assets execute in the consuming harness — not here — so governance never gets in the way of the work.

1

Publish

Author a skill, agent, MCP, or policy and submit it. Lugem validates metadata, scans for secrets, and routes it for review.

2

Review & version

Approve once, then ship immutable semantic versions. Dependencies resolve to a closure, and provenance travels with every release.

3

Consume everywhere

Pull into Claude Code, Cursor, Copilot, and CI with lugem pull and a lockfile — the same governance, audit trail, and version pin, whichever harness receives it.

The spine

A registry, hardened for the enterprise

Immutable semver and a dependency closure underneath; provenance, discovery, auth, and audit on top. Everything a package registry needs, plus the governance one demands.

Provenance & trust

Publisher identity, CODEOWNERS, and “verified by platform team” signals on every card.

Discovery at scale

Full-text search, trending, curated collections, and namespaces across your whole catalog.

Enterprise auth

Email, GitHub, Google, SSO (OIDC/SAML), passkeys, and TOTP, with Owner / Admin / Member roles.

Streaming audit

Append-only, per-org hash-chained events with search, export, retention, and SIEM delivery.

Bring-your-own storage

GitHub/Git, S3-compatible, HTTPS/artifact, and AWS Agent Registry backends over one interface.

Thin lugem CLI

A client over the same server APIs — every guardrail applies, with no new capability to govern.

Security & governance

Trust that is legible, and provable

lugem is multi-tenant by design and least-privilege by default. Your content stays in your infrastructure; every publish, approval, and pull is an auditable event you can stream to your SIEM.

Talk to our team

Deny-by-default authorization

Every request is authorized against org, group, and role before any asset is returned.

Per-org isolation, end to end

Tenants are isolated by org_id across the data plane, each with its own audit hash-chain.

Data residency you control

Content bytes never leave your storage backend — Lugem persists only metadata and a pointer.

Tamper-evident audit

Append-only, hash-chained events: verifiable, exportable, and streamable to your SIEM.

Pricing

Capacity-based pricing, no storage surprises

Pay for published assets, seats, and pulls — never for storage, because your content stays in your own backend. Start free and grow into it.

Free

Get started with the registry at no cost.

$0forever
  • Up to 25 published assets
  • Up to 5 seats
  • 1,000 pulls / month
  • Org / group / private visibility
  • Hash-chained audit log
Most popular

Team

For a growing org that needs more capacity.

Custombilled monthly
  • Up to 500 published assets
  • Up to 50 seats
  • 100,000 pulls / month
  • Everything in Free
  • SSO, passkeys & TOTP
  • SIEM audit delivery

Enterprise

Highest limits and org-wide governance.

Contact salesannual
  • Unlimited assets
  • Unlimited seats
  • Unlimited pulls
  • Everything in Team
  • Bring-your-own storage adapters
  • Priority support & onboarding

Every plan includes group-scoped visibility, the hash-chained audit log, and the lugem CLI. Compare plans in detail →

Bring order to your AI building blocks

Stop reinventing the same skill in five places. Publish once, govern centrally, and let every harness pull from one trusted source.