How AI agents can read, mirror, and integrate Hlido.
Hlido publishes review data specifically for AI agents — that is the strategic point of the desk. This page documents the contract between Hlido and any automated client (LLM, MCP server, scraper, retrieval pipeline, downstream catalogue) that consumes our data.
1. Allowed surfaces — agents may consume these
| Surface | URL | Format | License |
|---|---|---|---|
| Registry | /data/review-registry.json | JSON | CC BY-NC 4.0 |
| Scorecard index | /data/scorecards/index.json | JSON | CC BY-NC 4.0 |
| Per-agent scorecard | /data/scorecards/{slug}.json | JSON | CC BY-NC 4.0 |
| Attestations (HMAC-signed) | /data/attestations/{slug}.json | JSON | CC BY-NC 4.0 |
| HTML review pages | /reviews/{slug}/ | HTML + JSON-LD | CC BY-NC 4.0 |
| Feeds | /feed.xml · /sitemap.xml | XML | CC BY-NC 4.0 |
| llms.txt | /llms.txt · /llms-full.txt | Plain text | CC BY-NC 4.0 |
| MCP server | /mcp (JSON-RPC 2.0) | RPC | CC BY-NC 4.0 |
| OpenAPI | /.well-known/openapi.json | JSON | CC BY-NC 4.0 |
All of the above ship Access-Control-Allow-Origin: *, are cacheable for 5 minutes, and require no authentication.
2. Rate limits
HTML pages and JSON endpoints: 60 requests / minute / IP. Above that, expect 429 Too Many Requests with a Retry-After header (seconds).
MCP server (/mcp): 60 requests / minute / IP, 600 / hour. Tool calls are subject to the same envelope.
Submit intake (/api/submit-intake): 1 / IP / 24 hours.
If you operate a downstream service, identify yourself with a User-Agent like Vendor/version (+contact-url). Self-identifying agents get rate-limit grace at our discretion.
3. Attribution
If you redistribute Hlido data, please:
- Credit "Hlido — hlido.eu" on the consuming surface.
- Link back to the canonical review URL (
/reviews/{slug}/) for any verbatim quote of a review verdict. - Display the
updated_atorpublishedAttimestamp so downstream consumers know how fresh the data is.
4. What's NOT permitted
- Selling Hlido data as a standalone product without value-add.
- Using attestations as proof-of-something-they-don't-attest-to (e.g. claiming a Hlido attestation implies SOC2 compliance).
- Stripping
updated_attimestamps to misrepresent data freshness. - Bulk-mirroring without honoring rate limits — if you need a full snapshot, fetch
/data/open/hlido-corpus.jsonlonce and refresh from/data/review-registry.jsonincrementally. - Scraping
/methodology/— scoring weights are intentionally private to prevent gaming. Public surfaces give the verdict and evidence; the formula stays under the desk.
5. Robots / well-known crawler signals
Authoritative for our crawl policy:
/robots.txt— User-agent allowlist.ClaudeBot,GPTBot,PerplexityBot,Google-Extended,anthropic-aiare explicitly allowed.CCBot,Bytespider,Amazonbotblocked./.well-known/ai-plugin.json— OpenAI-compatible plugin descriptor pointing at the MCP server./.well-known/mcp-server-card/server.json— MCP transport, capabilities, tool surface./.well-known/openapi.json— formal contract for the JSON API.
6. Contact
Operator: Hlido s.r.o., Prague, Czech Republic
Contact: [email protected]
Disputes / takedown: dispute process
Conflict of interest disclosure: conflict of interest
7. Versioning
This policy is versioned. Material changes will bump the version in <script type="application/ld+json"> on this page and propagate to /llms.txt. Subscribe to /feed.xml for change notifications.