Hlido Agent Reviews — Claude Code MCP server: agent trust checks
Lets your agent look up trust scores, tier verdicts and claim audits for other AI agents. Useful when picking a ready-made agent for a job.
claude mcp add --transport http mcp https://hlido.eu/mcpWhat Hlido Agent Reviews does
The server exposes a registry of independent AI-agent reviews. Given a slug, it returns a 0–100 score, a tier and a verdict; given a single claim, it returns a PASS/FAIL result with evidence. Two agents can be compared side by side, and semantic neighbours of an agent are one call away.
Nine tools ship with it: trust_check, find_trusted, verify_claim, compare_agents, get_scorecard, find_similar_agents, submit_agent, report_review_issue and request_quick_audit, plus discovery and metadata helpers. It reads the same public JSON published under hlido.eu/data/*. Scoring weights, grader assertions and editorial drafts stay out of reach. Calls are anonymous JSON-RPC; no auth, no tracking.
The hosted worker lives at https://hlido.eu/mcp, so nothing has to be deployed. Self-hosting means a standard Cloudflare Worker: copy wrangler.toml.example to wrangler.toml, fill in the account id, run npx wrangler deploy. Docker skips the Cloudflare account entirely — the container runs the worker on local workerd through wrangler dev, with the MCP endpoint at http://localhost:8080/ (GET for server info, POST for JSON-RPC). KV cache and Vectorize bindings are optional; without them the worker fetches the public JSON directly.
What it helps with
- check whether an agent is worth trusting, by slug
- find a vetted agent for a specific need
- confirm or debunk a claim an agent makes about itself
- compare two agents on scores and verdicts
- flag a problem with a published review
Example prompts for the agent
- Use hlido to check if this agent is trustworthy
- Find me a trusted agent for document work through hlido
- Compare these two agents using their hlido scorecards
How to install Hlido Agent Reviews
Platform CLI: remote server
claude mcp add --transport http mcp https://hlido.eu/mcpPlatform CLI: remote server
codex mcp add hlido-agent-reviews --url https://hlido.eu/mcpConfig file
{
"mcpServers": {
"hlido-agent-reviews": {
"url": "https://hlido.eu/mcp"
}
}
}Config file
{
"servers": {
"hlido-agent-reviews": {
"type": "http",
"url": "https://hlido.eu/mcp"
}
}
}Platform CLI: remote server
gemini mcp add -t http hlido-agent-reviews https://hlido.eu/mcpStep by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Hlido Agent Reviews
How do I install Hlido Agent Reviews in Claude Code?
Run claude mcp add --transport http hlido-agent-reviews https://hlido.eu/mcp. The hosted server is already running, so there is nothing else to set up.
What does Hlido Agent Reviews do?
It gives your agent access to a registry of AI-agent reviews: score, tier, verdict, per-claim audits and comparisons. The data comes from public JSON.
Do I need an API key or an account?
No. The server answers anonymous JSON-RPC calls with no auth and no tracking. Daily per-tool counters are the only telemetry.
