For agents · For agents-of-agents · For humans

Connect to Hlido from anywhere.

Every public surface of Hlido in one place. Run the CLI in your terminal, gate PRs with the GitHub Action, plug 13 MCP tools into Claude / Cursor / Cline, sideload the browser extension, grab the corpus from HuggingFace, fork the public mirror on GitHub, or call the JSON API directly. All free, no keys required for read access.

See all 8 developer surfaces Open the MCP playground

Distribution

Where Hlido is published.

Every review ships to every surface below on every publish. You should never have to ask "where do I read Hlido" — pick the channel that fits your workflow.

Live

hlido.eu

Canonical home. Every review, the methodology, badge offer, and academy. C2PA-signed evidence on every page.

Browse reviews →

Live

Hlido MCP server

Native Model Context Protocol endpoint. 13 tools for agent-to-agent trust check, semantic search, claim verification, audit-on-demand, advisory subscribe, verdict explain, and task-based recommend. JSON-RPC 2.0, no auth required.

Try the playground →

Live

Hlido CLI on npm

npx @hlido/cli check <slug> — terminal-native scorecards. Subcommands: check, search, compare, tier. MIT licensed, source on the public mirror.

CLI quickstart →

Live

hlido-gate (GitHub Action)

Composite action that fetches the latest signed Trust Attestation, posts a sticky PR comment with score / tier / last-tested, and gates merges below a threshold. Drop-in 30-line YAML.

GitHub Action setup →

Live

Browser extension

Manifest v3 sideload for Chrome / Firefox / Edge. Surfaces the Laddoo Score, tier, and last-tested date on every reviewed-agent vendor page. Source on the public mirror.

Install the extension →

Live

Trust Attestations

HMAC-SHA256-signed JSON per scored agent. Static, per-slug, schema-versioned. Designed for agent-to-agent trust verification and CI gates.

Attestation schema →

Live

Open data dump (CC-BY-4.0)

Full corpus as JSONL. Permissive license, integrity checksum, manifest, quarterly refresh cadence. Built for academic citation and AI-training corpora.

Open data landing →

Live

JSON API

Per-agent sanitized scorecards with full claim-vs-evidence detail. Stable schema v1.0 — built for machine consumption by other agents.

Open registry JSON →

Live

HuggingFace dataset

hlido-eu/agent-benchmark — sanitized public-safe records appended on every publish. JSONL, ready for LLM training-time ingest.

Open on HuggingFace →

Live

Public GitHub mirror

The corpus, an llms.txt, and the MCP server card — all in hlido-public. Fork freely. Methodology weights stay private; everything else is on this repo.

Open on GitHub →

Live

Smithery

Listed as @ankitkapur1992/hlido. Auto-publishes the server card on every R3 publish — observability, logs, and distribution to thousands of MCP-aware agents.

View Hlido on Smithery →

Pending

awesome-mcp-servers

The community-curated MCP server list. A Hlido entry is auto-PR'd on every server-card change.

See the list →

Live

Social — X, LinkedIn, Facebook

Every published review gets a founder-voice post on each channel via Buffer. Cadence is spaced — no flood.

Follow on X →

13 tools, no auth, JSON-RPC 2.0

What the Hlido MCP server can do.

Drop these into any MCP-aware client (Claude Desktop, Cursor, Cline, Continue, Goose, the Anthropic SDK). Hit tools/list at https://hlido.eu/mcp to enumerate. All public, no key needed. The v2 trio (subscribe / explain / recommend) shipped 2026-05-09.

Tool	What it does	Use it when
trust_check	Returns the Laddoo Score, tier, blockers, and verdict for one agent.	You have a slug or URL and need a single-call trust answer.
get_scorecard	Full sanitized claim-vs-evidence scorecard with every command, exit code, and source surface citation. Schema v1.0.	You're an agent doing pre-flight evaluation before delegating work.
find_similar_agents	Semantic search over the corpus. Free-text in → top-k matches with Laddoo scores out.	You have a task description and want Hlido's recommendation.
find_trusted	Substring + token search across reviewed agents.	You know roughly what you're looking for and want a list.
verify_claim	Checks whether a Hlido review references a specific claim. Honest nulls when not tested.	You want to know "did Hlido test claim X for agent Y?"
compare_agents	Side-by-side comparison of up to 5 reviewed agents.	You're choosing between alternatives.
request_quick_audit	Queue an audit for an agent we haven't reviewed yet. Returns the future scorecard URL + ETA.	You encounter an unknown agent and need a verdict.
report_review_issue	Flag a stale verdict, wrong claim, or broken link. Routes to the dispute-retest queue.	You can prove a current Hlido review is incorrect.
submit_agent	Submit a brand new AI agent for Hlido's daily intake.	You want a slow-queue review (next daily cycle).
subscribe	Register an advisory watch on a slug. Returns a watch handle the calling agent can poll for score-change notifications.	You want to be told when a reviewed agent's score moves.
explain	Returns the verdict reasoning behind a Laddoo Score — which claims passed / failed and why.	You need a justification you can show a buyer or a downstream agent.
recommend	Free-text task description in → ranked list of reviewed agents that fit out, with score and reasoning.	You have a task and want the best-scoring agent for it.

2-minute setup

Connect your MCP client.

Same JSON snippet works in every modern MCP client. Replace ~/.claude/... with the equivalent path for your client.

Claude Desktop / Claude Code

Edit ~/.claude.json (or open Claude Code → settings → MCP servers).

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

Cursor

Cursor → Settings → MCP → add server.

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

Cline / Continue / Goose

Each client has its own MCP config UI; the URL is what matters.

# MCP endpoint
https://hlido.eu/mcp

# Health-check
curl https://hlido.eu/mcp \
  -X POST -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

No MCP? Just curl.

Direct JSON endpoints.

Read-only. Free. CDN-cached at the Cloudflare edge. Same data the MCP server reads.

GET https://hlido.eu/data/review-registry.json — full sanitized review index
GET https://hlido.eu/data/scorecards/{slug}.json — per-agent claim+evidence (e.g. aider, gumloop)
GET https://hlido.eu/.well-known/mcp-server-card/server.json — full MCP server card with tool schemas
GET https://hlido.eu/llms.txt — LLM-friendly discovery hint
GET https://hlido.eu/llms-full.txt — full review corpus for AEO ingest
GET https://hlido.eu/feed.xml — RSS 2.0 of published reviews
GET https://images.hlido.eu/reviews/{slug}/{run-id}/{file}.png — C2PA-signed proof screenshots

Build with Hlido

Test the MCP server in 30 seconds.

Open the playground. No login, no setup. Hit the tools live and watch the JSON-RPC responses.

Open the playground Submit your agent