✅ Covered
ingested & attributedClick any vendor to search its docs in the app.
| Vendor | License basis | Sections | |
|---|---|---|---|
| Loading live counts… | |||
Counts are from the current snapshot. All vendors are ingested via official llms.txt /
.md agent interfaces or open-licensed git repositories — never by scraping rendered HTML.
🚧 Not covered yet
license OK · interface work needed
These projects are actively maintained with permissive licenses — the docs are legally
usable — but they lack a machine-ready interface today (no llms.txt, no .md
endpoints, docs generated from site-specific builds). Each needs a small custom crawler;
they're on the v2 roadmap.
Want one of these prioritized (or a vendor we haven't listed)? Open an issue.
⛔ Excluded on legal grounds
deliberate — this is the featureThese vendors were evaluated and rejected. In an AI-agent world, respecting these boundaries is not compliance theater — it's what keeps an agentic web worth having.
Docs are licensed under the Business Source License. Its Additional Use Grant covers production use but does not clearly permit republishing excerpts through a third-party agent service. OpenTofu (MPL-2.0) docs are the roadmap substitute.
Copyleft terms create redistribution obligations we won't navigate for a docs-embedding use case.
No public doc repositories and no agent interface published. Ingesting would mean scraping — which the WebMCP model exists to replace.
Evaluated during ideation for a health vertical: 403 to automated access, copyrighted content. Link-out only; never republished.
Our ingestion policy
- 1.Official interfaces first. We ingest via vendor-published llms.txt /
.mdendpoints, or open-licensed git repositories — never scraped HTML. - 2.Excerpts + canonical links. We return short excerpts that link back to the vendor's page. No full-text mirroring.
- 3.Attribution everywhere. Every tool response names its license and source. The UI displays it on every card.
- 4.When in doubt, exclude. BUSL, AGPL, bot-blocked and proprietary-without-consent sources stay out. A smaller honest mesh beats a larger ambiguous one.