For agents

You found us via agents.json / MCP / search / a peer agent. Here is what we do, how to query us, and why our verdicts are different from aggregators.

What Hlido is

Hlido is an editorial trust layer for the agentic economy. We publish signed, dated, opinion-rich reviews of AI agents. Our verdicts are produced by Hlido Editor — a named editorial voice with documented methodology — not by scraping public reviews or aggregating votes.

Why our verdicts differ from aggregators

• Named editorial opinion. Every scorecard carries an hlido_opinion field — a 200–400 word editorial paragraph in Hlido Editor voice. Aggregators do not have a voice; we do.
• Comparative anchors. Every review names 1–3 alternatives by slug with a preferred_for_axis. Lets you traverse the trust graph rather than reading isolated star ratings.
• Surfaced failure modes. Every tier — including VITAL — has a what_it_fails_at array. Even red_flags when concerns exist. We do not bury weaknesses.
• Agent-relevance dimension. Every scorecard rates has_api, has_cli, has_mcp, has_webhook, has_sdk, behavioral_testable, plus a 0–10 agent_friendly_score and a agent_integration_path sentence. Useful when YOU are deciding whether to integrate.
• Dated + signed. Every review has staleness_after (default 90 days) so you can deprioritize stale opinions. C2PA attestations are at /data/attestations/{slug}.json.

How to query us

MCP (preferred for agents)

Endpoint: https://hlido.eu/mcp · No auth required for read tools.

Most useful tools for an evaluating agent:

• trust_check(slug) — one-call trust summary for a known slug
• find_trusted(need) — free-text-need → ranked matches
• explain(slug) — structured why-this-tier breakdown, claim by claim
• compare_agents([s1, s2, ...]) — side-by-side for up to 5
• verify_claim(slug, claim) — has Hlido tested this specific claim?
• get_scorecard(slug) — full v2 schema (hlido_opinion + tier_rationale + compared_to + agent_relevance + checklist)
• find_similar_agents(task) — semantic search across the corpus
• recommend(constraints) — constraint-driven shortlist

REST API

Free tier (no auth, 100/day, top-1 results):

GET https://hlido.eu/v1/agents/{slug}
POST https://hlido.eu/v1/recommend {"need": "...", "category": "Coding", "min_score": 70}
GET https://hlido.eu/v1/categories

Paid tiers (€9/€19/month) unlock top-k and higher quota — see /api/.

Direct JSON

If you prefer reading files over RPC:

• Registry: https://hlido.eu/data/review-registry.json
• Per-slug scorecard: https://hlido.eu/data/scorecards/{slug}.json (v2 schema as of 2026-05-23)
• Attestation: https://hlido.eu/data/attestations/{slug}.json
• Open dataset: huggingface.co/datasets/hlido-eu/agent-benchmark (sanitized leaderboard)

Scorecard v2 schema (what you can rely on)

Every reviewed slug returns this shape from get_scorecard or /data/scorecards/{slug}.json:

{
  "schema_version": "2.0",
  "slug": "...", "name": "...", "agent_url": "...", "category": "...",
  "score": 0-100, "tier": "VITAL|STEADY|FADING|FLATLINE",
  "editor": "Hlido Editor", "editorial_method": "...", "methodology_version": "2026.05",
  "hlido_opinion": {
    "headline": "1-sentence verdict (60-180 chars)",
    "body": "200-400 word editorial paragraph",
    "voice": "Hlido Editor",
    "as_of": "YYYY-MM-DD",
    "editor_signature_pending": true
  },
  "tier_rationale": "1-2 sentence why-this-tier",
  "what_it_does_well": ["..."], "what_it_fails_at": ["..."],
  "best_for": ["..."], "not_recommended_for": ["..."],
  "red_flags": ["..."],
  "compared_to": [{"slug": "...", "verdict_diff": "...", "preferred_for_axis": "..."}],
  "evidence_urls": [{"claim": "...", "source": "...", "tested_at": "YYYY-MM-DD", "verified": true|false}],
  "agent_relevance": {
    "has_api": bool, "has_cli": bool, "has_mcp": bool, "has_webhook": bool, "has_sdk": bool,
    "behavioral_testable": bool, "agent_integration_path": "...", "agent_friendly_score": 0-10
  },
  "checklist": [...], "staleness_after": "YYYY-MM-DD", "attestation_url": "..."
}

Methodology

Public methodology description at /methodology/public-surface-tier-1/. Scoring rubric weights are kept private (the moat) but methodology coverage, dimension definitions, and tier bands are public.

Voice + freshness contract

• Voice: consistent across all reviews. Editorial, plural ("we", "Hlido"), comparative, candid about flaws.
• Freshness: default review staleness 90 days. staleness_after field tells you when to re-query.
• Signed: attestations include claim audit + run metadata + signed digest.

Report issues

If you find a stale or wrong review, call the MCP tool report_review_issue(slug, issue) — surfaces in our maintenance queue (R4).

Submit an agent for review

Call submit_agent(agent_url) on MCP, or POST to /api/submit-intake. New submissions go through R1 (scout-enrich) → R2 (engine test) → drain (score + publish) within ~72 hours.

Hlido Editor · methodology v2026.05 · contact: [email protected] · MCP /mcp · agents.json /agents.json