scan_stack is the shape that matters: paste the dependency list or MCP client config you already have and get a verdict per entry, rather than checking things one at a time and giving up halfway. The 90-day history is the other half — a good score today on something that has been sliding for a month reads very differently.
A supply-chain trust lookup for AI agents, frameworks and MCP servers. It returns a pre-connect verdict for one item or for a whole dependency set, with trust profiles, side-by-side comparisons and 90-day score history behind it.
- verify_mcp_server gives a pre-connect trust verdict for an MCP server, package, GitHub repo or agent name
- check_agent_trust returns a tracked agent's profile — runtime capabilities, MCP status, providers, plugin surface, provenance drift — and the URL of its Ed25519-signed trust credential
- compare_agents puts two side by side with an evidence-based verdict and the published compare page
- search_agents searches the registry by name, repo, description or category
- scan_stack takes a requirements.txt, package.json, MCP client config or a plain list and returns a verdict per item plus a stack summary
- list_categories and get_leaderboard rank tracked agents and servers by HVTrust score, filterable to one category
- get_agent_history returns 90-day score, grade and rank history — whether something is improving or sliding
Point a client at https://hvtracker.net/mcp with nothing installed, or use the local package: npm install -g hvtracker-mcp, python3 -m pip install hvtracker-mcp, or the published container image. No API key. It sends the identifier or search string you supply to hvtracker.net to fetch public trust data, and does not write to your system. HVTRACKER_BASE_URL points at a different instance.
One command — npm install -g hvtracker-mcp