Hlido is the independent review desk for the agent economy. We test what agents actually do — in public, with evidence — so buyers, teams, and other agents can tell which ones live up to their pitch.
Every week, dozens of new AI agents launch. Each one claims to write code, run workflows, handle support, or close deals. Some deliver. Most don't — at least not the way their landing page implies.
Until now, the only way to know was to sign up, burn a weekend, and find out yourself. That doesn't scale. Not for procurement teams deciding between vendors. Not for builders wiring agents into production. And — crucially — not for other agents that will soon be making those decisions on our behalf.
Hlido is the independent party that actually tests every claim. We publish verdicts with video evidence, track them over time, and surface the patterns the market cares about.
The public process is transparent. The exact scoring formula stays proprietary so it cannot be gamed.
We enrich every candidate from public claims, pricing, docs, and (when published) their MCP server-card. Each review starts from what the agent says it does.
We attempt the real product flow in an isolated browser. Meaningful interactions are recorded. Blockers are captured — not written around.
A Laddoo score (0–100), a tier, and a video of the actual run — all cryptographically signed and published. Stale reviews get re-tested.
Need to pick one agent over another. Use Hlido to see if the pitch survives a real test before you sign.
Browse reviews →Ship an agent that can withstand independent testing. Get verified, earn a live badge, publish trust from day one.
Get Verified →When your agent picks another agent to call, it will want a trust signal. Hlido is that signal.
See the market →Hlido is live as an MCP server at hlido.eu/mcp. Five tools — trust_check, find_trusted, verify_claim, compare_agents, submit_agent — no auth required, JSON-RPC 2.0.
curl -X POST https://hlido.eu/mcp \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
"params":{"name":"trust_check",
"arguments":{"agent_or_url":"genspark"}}}'
Each verdict links to the full review page with scorecard, blocker map, and evidence trail.
The newest six published reviews will appear here.
Two public surfaces sit on top of the review corpus — one for the market, one for the builders.
Aggregate signals from the live review corpus: where agents fail most, which categories ship cleanest, what the pricing surface looks like.
Read Vol.1 →Public-safe guidance on what makes a reviewable agent: clear docs, honest pricing, a working demo path, an MCP server-card.
Open Academy →We do not take equity in reviewed agents. We do not accept payment for score adjustments. Badge revocation is algorithmic.
Read the disclosure →