Hlido

What is Hlido?

Hlido is an independent review platform that provides evidence-backed evaluations of AI agents. It offers C2PA-signed proof, machine-readable scorecards, and a consistent Laddoo Score (0–100) for every reviewed agent. It is designed for buyers, teams, and other agents that need to verify claims about AI agents. Hlido also exposes a live MCP server with 9 tools for querying reviews and scores.

How to use Hlido?

Use the MCP server at https://hlido.eu/mcp via JSON-RPC 2.0. No authentication is required, and it is free to call. Tools include trust_check, find_similar_agents, get_scorecard, request_quick_audit, verify_claim, and compare_agents. You can also access the public mirror repository on GitHub to use the scorecard JSON files directly from your own scripts, agents, or analysis pipelines.

Key features of Hlido

• Independent, evidence-backed reviews of AI agents
• Consistent Laddoo Score (0–100) across all agents
• C2PA-signed screenshots and terminal casts for proof
• Live MCP server with 9 JSON-RPC 2.0 tools
• Structured machine-readable scorecards in JSON format
• Independent MCP-server safety register with per-flag evidence

Use cases of Hlido

• Procurement teams evaluating and comparing AI agent vendors
• Builders wiring agents into production and needing verified claims
• Other AI agents autonomously making decisions based on reviews
• Auditing MCP servers for safety (tool-poisoning, dangerous capabilities)
• Tracking agent performance over time with consistent scoring

FAQ from Hlido

What is the Laddoo Score?

The Laddoo Score is a consistent score from 0 to 100 derived from a fixed evaluation framework. Tier mapping: VITAL ≥ 90, STEADY ≥ 70, FADING ≥ 40, and FLATLINE < 40.

How is the evidence verified?

Screenshots and (where applicable) terminal casts are cryptographically signed with C2PA, so the captured proof can be verified as un-tampered.

Does the MCP server require authentication?

No. The MCP server at https://hlido.eu/mcp is free to call and requires no authentication. It uses JSON-RPC 2.0.

Where can I access the review data?

Reviews are mirrored in this public repository. Machine-readable scorecards are available as JSON files at scorecards/{slug}.json. The data is also indexed on Smithery, HuggingFace, and the awesome-mcp-servers list.

What is the MCP-server safety register?

The register scores MCP servers based on tool-poisoning, dangerous-capability, and auth-posture red-flags, with published per-flag evidence. Classifications are SAFE, CAUTION, RISKY, or DANGEROUS.