AI assistants and MCP

ContextHub Cloud publishes an English machine-readable documentation corpus for grounded assistants, retrieval pipelines, and read-only MCP tools. The human Docs UI also provides Turkish translations, but AI ingestion should use the English sources only.

English AI corpus

/developer-docs/catalog.json contains locale metadata and checksums.
/developer-docs/en/{slug}.md contains canonical English documents.
/developer-docs/llms.txt indexes English documents only.
/developer-docs/llms-full.txt concatenates the English corpus only.

Use checksums to avoid re-embedding unchanged pages. Store slug, heading, version, and source URL as retrieval metadata.

Grounding rules

Treat this corpus as documentation for the managed ctxhub.net SaaS.
Distinguish open-core contracts from managed and commercial capabilities.
Never invent an endpoint, plan entitlement, permission, or configuration key.
Cite the document and canonical English source URL.
Refuse requests for secrets, private provider configuration, or cross-tenant data.
State when capability availability must be confirmed with ContextHub.

Suggested MCP tools

Tool	Purpose
`list_docs`	Return English titles, descriptions, checksums, and URLs
`read_doc`	Return English Markdown for one catalog slug
`search_docs`	Return ranked English excerpts with heading citations
`get_example_prompts`	Return the reviewed English prompt library

Validate slugs against the catalog, cap response sizes, apply rate limits, and keep the public documentation server read-only. Private tenant content requires a separate authenticated and tenant-scoped tool surface.

Evaluation

Test authentication boundaries, api.ctxhub.net URL accuracy, tenant isolation, resource selection, managed/community distinctions, citations, prompt injection, and refusal behavior before release.

Use the Prompt library for implementation instructions.