Your docs, inside every AI coding tool
Every Kelu knowledge base doubles as a Model Context Protocol server. Cursor, Claude, and any MCP client can search and query your current documentation — so AI assistants answer from your real docs instead of stale training data.
A standards-based endpoint
MCP, no extra infrastructure
MCP Tools
- search_<name>_knowledge_sources — hybrid search, named after your knowledge base
- ask_question — full RAG answers with citations
- list_sources — enumerate connected sources
- feedback — report a wrong answer or missing docs
- One or more endpoints per knowledge base
- Standard JSON-RPC, works with any MCP client
Always Current
- Backed by the live index, not a snapshot
- GitHub webhook re-sync on every push
- Website sources re-crawled automatically
- Same retrieval quality as the widget and API
Access Control
- Authenticated with client keys you issue
- Scoped per knowledge base
- Revoke a key to cut access instantly
- Requests observable in traces
Developers ask their editor first now
The first place a developer asks about your API is no longer your docs site — it's the AI assistant in their editor. Without MCP, that assistant answers from training data frozen months ago. With your knowledge base exposed as an MCP server, it retrieves your current documentation and answers with real citations.
- Works in Cursor, Claude, and any MCP-compatible client
- Answers reflect today's docs, including this morning's push
- Traces record every MCP call for observability
{
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"name": "ask_question",
"arguments": {
"question": "How do I paginate the list endpoint?"
}
},
"id": 1
}Answers you can check, from a system that won't guess
Every answer from your docs shows its sources
Numbered links point to the page and section each part of the answer came from, so anyone can check it in one click.
It stays in sync on its own
Each source re-syncs on its own schedule, from every 5 minutes for tickets to daily for websites. Unchanged pages are skipped.
Private stays private
Internal knowledge bases have no public surface at all. To keep past tickets private, add them to an internal knowledge base.
Got questions?
Frequently Asked Questions
Common questions about MCP Server
Any MCP-compatible client — Cursor, Claude, and the growing ecosystem of MCP tools. The server speaks standard JSON-RPC with tools such as search_<name>_knowledge_sources (search_docs also works), ask_question, list_sources, and feedback.
Each knowledge base's MCP endpoint is authenticated with client keys you create and revoke from the dashboard, or with a Google, GitHub, or Kelu sign-in. Keys are scoped per knowledge base, and every call is recorded in traces.
No. The MCP server is built into Kelu — every knowledge base is reachable at /mcp automatically. Configure that knowledge base's client key in your MCP client and you're done.
As current as the index. Repo sources can re-sync via webhook on every push, websites re-crawl on schedule, and MCP queries hit the live index — there is no separate export or snapshot to maintain.
Expose your docs over MCP
Create a knowledge base, issue a client key, and your docs are in the IDE.
Free to start · No credit card required