AI & Agent Tools
LPM.dev Registry uses AI for publication security review, public package discovery, and structured context for coding agents. It does not automatically analyze every package published to the Registry.
Private packages remain outside automatic and background AI processing. The public-package workflows begin only when an owner chooses Pool or Marketplace distribution.
Two Separate Package Pipelines
| Pipeline | When it runs | Effect |
|---|---|---|
| Publication security review | A package enters Pool or Marketplace, and whenever a new version is published there | Blocks that version from consumers until it is approved |
| Package enrichment | After an eligible public-distribution version is approved | Adds summaries, search embeddings, API documentation, usage context, and compatibility data |
The publication review is mandatory. Enrichment is asynchronous and does not decide whether a version may be installed.
See AI Analysis & Metadata for the complete version lifecycle and privacy model.
Package Discovery
Approved Pool and Marketplace packages can receive source-grounded metadata such as:
- a concise description, capabilities, use cases, and tags
- guidance about when a package is or is not a good fit
- a small quick-start example
- structured API documentation and an agent-oriented usage guide
- compatibility signals extracted from package metadata
- an embedding used for semantic search
These fields support package pages and intent-based search. Availability varies by version: newly approved metadata can still be processing, and a package without analyzable source may not receive every artifact.
MCP Server
The LPM.dev Registry MCP Server gives compatible coding agents structured tools for package search, package context, Marketplace access, Pool statistics, Registry documentation, and LPM CLI-backed project actions.
Use it when an agent needs current Registry data instead of relying on model training data or asking you to paste command output.
Agent Skills
There are two related but separate skill workflows:
- Package-published Agent Skills are written by package authors and teach coding agents how to use a specific package. LPM CLI can install them with the dependency.
- The LPM Guide is a standalone skill for authors working with LPM.dev Registry publishing, quality, distribution, pricing, CI, Swift packages, and package-skill creation.
See Package Agent Skills for the package format and LPM Guide for the standalone authoring assistant.
Website AI Chat
After signing in, open AI chat from the chat button in the lpm.dev header. It can search packages, inspect package details and access models, compare quality information, and browse publishers.
Public discovery uses public Registry metadata. Asking about a private package is an explicit authenticated action with a separate disclosure before relevant private metadata is sent for that response. See the Privacy Policy for details.
