MCP server — tool reference

Endpoint: https://hlido.eu/mcp — JSON-RPC 2.0 (Streamable HTTP). No authentication for read tools. Listed on the Official MCP Registry, Smithery, mcp.so and Glama.

Looking for Hlido's safety scores of third-party MCP servers instead? That's the MCP Trust register (human view). This page documents Hlido's own MCP server.

Add to your client#

{
  "mcpServers": {
    "hlido": { "url": "https://hlido.eu/mcp" }
  }
}

Works in Claude Code (claude mcp add --transport http hlido https://hlido.eu/mcp), Cursor, Cline, Continue, Goose — any MCP-capable client.

Tool reference#

Tool	Arguments	Returns
`trust_check`	`agent_or_url`, `use_case`?	One-call trust summary: score, tier, headline verdict, staleness.
`find_trusted`	`need`, `min_tier`?, `limit`?	Ranked reviewed agents matching a free-text capability need.
`verify_claim`	`agent`, `claim`	Whether Hlido has independently tested that specific claim, with evidence.
`compare_agents`	`slugs` (≤5)	Side-by-side scores, tiers, strengths and failure modes.
`get_scorecard`	`slug`	Full v2 scorecard — see the schema reference.
`explain`	`slug`, `dimension`?	Structured why-this-tier breakdown.
`recommend`	`constraints`	Constraint-driven shortlist (category, must-have surfaces, min score).
`find_similar_agents`	`description`, `top_k`?, `min_score`?	Semantic matches across the corpus.
`get_incidents`	`slug`?, `severity`?, `category`?, `limit`?	Published reliability incidents.
`get_behavioral_trace`	`slug`, `spec_version`?	Verified behavioral test runs where available.
`commerce_check`	`agent_or_url`	Agentic-Commerce Readiness signals (interface surface, delegation safety).
`scan_mcp`	`server`, `requester`?	Safety scan of a third-party MCP server — see the MCP Trust register.
`verify_transparency`	`url`	Live EU AI Act Article 50 public-surface probe — works for agents we have never reviewed.
`market_pulse`	`category`?	Agent Market Pulse — corpus-level movement, with its own coverage statement.
`intel_query`	`sector`?, `geo`?, `compliance`?, `protocol`?, `kind`?, `audience`?, `maturity`?, `text`?, `limit`?	Faceted query across the intel store.
`submit_agent`	`url`, `name`, `note`?, `email`?	Queues an agent for independent review (free).
`report_review_issue`	`slug`, `issue_type`, `detail`, `reporter`?	Files a correction ticket against a review.
`request_quick_audit`	`url`, `name`?, `why`?, `requester`?	Requests a fast surface audit (rate-limited per requester per day).
`subscribe`	`slug`, `channel`?	Change alerts for an agent's score.

Arguments are generated from the live server schema, so this table cannot drift from what https://hlido.eu/mcp actually accepts. ? marks an optional parameter. The server also accepts the obvious aliases for each required parameter (slug for agent_or_url, task for description, and so on) — but the names above are canonical, and a missing required parameter returns a self-correcting error naming it.

Example call#

curl -s https://hlido.eu/mcp -X POST -H 'content-type: application/json' -d '{
  "jsonrpc": "2.0", "id": 1, "method": "tools/call",
  "params": { "name": "trust_check", "arguments": { "agent_or_url": "aider" } }
}'

MCP vs raw JSON#

Use MCP when an agent decides at runtime (intent-shaped tools, no schema knowledge needed). Use the data endpoints when you want bulk data, caching control, or zero protocol overhead.