knowledgelib-mcp

MCP server for knowledgelib.io — gives AI agents access to 1,564 structured, cited knowledge units across 16 domains. Pre-verified answers with inline source citations, confidence scores, quality status, and freshness tracking.

Quick start

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "knowledgelib": {
      "command": "npx",
      "args": ["knowledgelib-mcp"]
    }
  }
}

MCP over HTTP (no install needed)

POST https://knowledgelib.io/mcp

Streamable HTTP transport, JSON-RPC 2.0, MCP spec 2025-03-26. Same 6 tools, works with any HTTP-capable MCP client.

Other MCP clients (Cursor, Windsurf, etc.)

npx knowledgelib-mcp

The server uses stdio transport and works with any MCP-compatible client.

Tools

query_knowledge

Search across all knowledge units by relevance. Returns confidence scores, token estimates, quality status, content previews, and related units.

batch_query

Search multiple topics in a single call. Shares a single catalog parse — more efficient than calling query_knowledge multiple times.

get_unit

Retrieve a specific knowledge unit as raw markdown with YAML frontmatter. Prepends quality warning if the unit has open issues.

list_domains

List all available knowledge domains with unit counts. No parameters.

suggest_question

Submit a question or topic request. Popular suggestions are prioritized for new unit creation.

report_issue

Flag incorrect, outdated, or broken content on a knowledge unit.

Tool Annotations

All tools include MCP spec 2025-03-26 annotations:

Read-only tools can be called in parallel by agents.

Environment variables

What you get

Each knowledge unit includes:

• Structured metadata — YAML frontmatter with confidence, sources, temporal validity, jurisdiction

• Quality status — verified, needs_review, or unreliable (based on open issues and confidence)

• Content previews — 150-char summaries without fetching full content

• Related units — knowledge graph edges (related_to, depends_on, alternative_to, often_confused_with)

• Inline citations from 5-8 authoritative sources

• Token estimates for context budget planning

• Batch queries — search up to 10 topics in one call

Why use this instead of web search?

REST API

The MCP server wraps the knowledgelib.io REST API. You can also use the API directly:

GET  https://knowledgelib.io/api/v1/query?q=best+earbuds&limit=3
POST https://knowledgelib.io/api/v1/batch
GET  https://knowledgelib.io/api/v1/units/{id}.md
GET  https://knowledgelib.io/api/v1/health
POST https://knowledgelib.io/api/v1/suggest
POST https://knowledgelib.io/api/v1/feedback

Full API docs: knowledgelib.io/api OpenAPI spec: /api/v1/openapi.json

License

CC BY-SA 4.0