AgentsHQ
A package manager and public registry that makes reusable AI-agent definitions discoverable, portable and installable across more than forty coding tools.
The problem worth solving
AI coding tools describe agents in incompatible files and directories. Useful definitions become trapped in one repository or one editor, while users lack a consistent way to discover them, inspect their provenance, assess risk and keep installations current.
What I built
AgentsHQ applies package-manager conventions to agent definitions. Its CLI installs from a source repository, detects supported coding tools, translates the canonical definition into their native formats and records the result in a lock file for repeatable updates and removal.
The public registry indexes repositories and individual agents, adds search and installation signals, and displays automated security analysis on detail pages. One command can install a full collection or a specific agent into the tools already present in a project.
What the product does
- One-command installation from a repository or a specific agent definition.
- Native-format output for at least forty-three AI coding IDEs and command-line tools.
- Lock-file tracking for repeatable updates, listing and clean removal.
- A searchable public registry with source provenance and installation signals.
- Layered static and language-model security analysis for indexed definitions.
Why the design matters
Portability is more useful than another proprietary agent format. AgentsHQ keeps source ownership with publishers while giving users one installation model, visible provenance and a safety signal across an ecosystem that otherwise changes tool by tool.
My role
Ciprian created AgentsHQ and leads its package model, command-line workflow, format translation, registry experience and security-analysis direction.