CLI, Python package, GitHub Action, plugin

Four ways to consume Hlido inside your own tooling. All free, all read the same public data.

CLI (zero-dep, Node 18+)#

npx @hlido/cli check aider
npx @hlido/cli search "code review"
npx @hlido/cli compare aider cursor

hlido audit — "npm audit for AI agents"#

npx @hlido/cli audit

Scans the current project (package.json, requirements.txt, MCP client configs — .mcp.json, Cursor, Claude Desktop) for AI-agent/MCP dependencies and prints a trust table: score + tier per reviewed agent, flags for FLATLINE scores and stale reviews, and a free-submit pointer for anything unreviewed. Exits non-zero when something is flagged, so it works as a CI gate too. --json for machine output.

Python (agent frameworks)#

pip install hlido

Trust tools (trust_check, find_trusted, verify_claim, get_scorecard) auto-register in environments that already have the framework: NVIDIA NeMo Agent Toolkit (via the nat.plugins entry point), LangChain, CrewAI. One package, no per-framework installs.

GitHub Action — hlido-gate#

Fail PRs when a dependency's trust score regresses:

- uses: ankitkapur1992-hlido/hlido-public/actions/hlido-gate@main
  with:
    slug: your-agent-dependency
    min-score: 70

Claude Code plugin#

/plugin install hlido@hlido
/hlido check aider

Bundles the MCP trust tools + /hlido:check + /hlido:recommend. Repo: github.com/hlido/hlido-plugin.

Browser extension#

See the Hlido score on vendor pages as you browse — Chrome / Firefox / Edge sideload: /extension/.