Blck Alpaca

唯讀存取 Blck Alpaca 的知識庫：SEO、GEO、AI 代理、行銷自動化。

文件

Blck Alpaca Knowledge Base MCP

Public MCP server for the Blck Alpaca knowledge base: SEO, GEO (generative engine optimization), AI agents and marketing automation, plus a website audit for AI-agent readiness. Read-only, no authentication, German, English and Slovak.

Listed in the official MCP Registry as at.blckalpaca/knowledge-base.

This repository holds metadata and the Claude plugin. The server runs as part of blckalpaca.at and its source is not published here.

Endpoint

POST https://blckalpaca.at/mcp

Transport: Streamable HTTP. No session is created — there is no Mcp-Session-Id and no GET stream. Every call is a single POST.

Install

Claude plugin with the server and a skill that uses it (see plugins/blck-alpaca):

claude plugin marketplace add blckalpaca-studio/knowledge-base-mcp
claude plugin install blck-alpaca@blckalpaca

Antigravity plugin, same server and skill (see plugins/blck-alpaca-antigravity):

agy plugin install https://github.com/blckalpaca-studio/knowledge-base-mcp/tree/main/plugins/blck-alpaca-antigravity

Server only, Claude Code:

claude mcp add --transport http blckalpaca-knowledge-base https://blckalpaca.at/mcp

Any client that reads a config file:

{
  "mcpServers": {
    "blckalpaca-knowledge-base": {
      "type": "http",
      "url": "https://blckalpaca.at/mcp"
    }
  }
}

Protocol revisions

The same URL serves two generations, so clients built on older SDKs keep working.

Revision	Client entry point	Required headers
`2026-07-28`	`server/discover`, `tools/list`, `tools/call` directly	`MCP-Protocol-Version`, `Mcp-Method`; on `tools/call` also `Mcp-Name`
`2025-11-25`, `2025-06-18`, `2025-03-26`	`initialize`, then `notifications/initialized`	none

An unknown version gets JSON-RPC error -32022 with the full list under data.supported.

Tools

Tool	Purpose	Required	Optional
`search_knowledge_base`	Full-text search over article title, short definition and main keyword. Returns URL plus a path triple for the full text.	`query` (string, min 2 chars)	`locale` (`de`\|`en`\|`sk`, default `de`), `limit` (integer 1–25, default 8)
`list_knowledge_categories`	Structure of the knowledge base: all categories with their topics, each with slug and URL. Entry point when no search term is known.	—	`locale`
`list_knowledge_topic_articles`	All published articles of one topic.	`categorySlug`, `topicSlug`	`locale`
`get_knowledge_article`	Full article as Markdown, plus short definition, key takeaways, sourced statistics and FAQ.	`categorySlug`, `topicSlug`, `articleSlug`	`locale`
`audit_website`	Checks a public website for AI-agent readiness and GEO (GET only, no login): score out of 100 across five weighted areas, evidence for every check, the three biggest fixes with a Knowledge Base guide where one exists, link to the full report.	`url`	`locale` (`de`\|`en`\|`sk`, default `en`)

Slugs are localized. Category and topic slug must come from the same locale as the request; if they do not match, the response says why. Content that is not translated into the requested language is omitted rather than silently replaced with German.

The path triple returned by search_knowledge_base and list_knowledge_topic_articles can be passed straight into get_knowledge_article.

Rate limit

240 requests per hour and IP; audit_website additionally at most 10 checks per website per hour and 60 in total. Every response carries RateLimit and RateLimit-Policy (IETF draft) plus the legacy RateLimit-Limit / -Remaining / -Reset triple. Over the limit the server answers HTTP 429 and JSON-RPC -32603 with data.retryAfterSeconds, plus a Retry-After header.

Using the content

Articles may be quoted and summarized with attribution to Blck Alpaca and a link to the article URL returned by the tool.

Contact

office@blckalpaca.at — Blck Alpaca e.U., Vienna, Austria