KlyHub Developer Documentation

Welcome to the KlyHub developer docs. KlyHub gives you one structured knowledge base for each of your companies — and lets you reach it from any AI client that speaks MCP (Model Context Protocol). Claude Desktop, Cursor, ChatGPT, or your own custom agent: same knowledge, no copy-paste, no re-prompting.

What KlyHub gives you

• A multi-tenant knowledge base — one workspace per company, isolated by row-level security so nothing leaks between tenants.
• A 4-phase AI-guided intake — KlyHub interviews you to build out the knowledge graph step by step, instead of asking you to fill in blank forms.
• A 5-layer ontology — Core, Market, Motion(s), Operations, and Memory. Every fact in your KB lives in exactly one layer so retrieval stays fast and explainable.
• A per-tenant MCP server — your AI clients connect to mcp.klyhub.com/v1/<your-tenant-slug> over OAuth 2.1 and get tools to query (and, with permission, modify) your knowledge.

Where to start

If you have not signed up yet, head to app.klyhub.com and create your first workspace. The intake will walk you through it.

If you already have a workspace, jump to the Getting Started guide. It takes you from "logged in" to "Claude Desktop is talking to my knowledge base" in five steps.

For deeper reference material, the sidebar groups every guide by topic:

• MCP setup — per-client setup for Claude Desktop, Cursor, ChatGPT, and custom agents.
• Intake methodology — how the 4-phase intake works, end to end.
• Ontology — what the 5 layers mean, and how to decide where a fact belongs.
• API reference — the full REST + oRPC surface at api.klyhub.com.

Why MCP, and why now

MCP (Model Context Protocol) is an open standard for connecting AI clients to external context and tools. We picked it because it's the only protocol where Claude, Cursor, and ChatGPT all converge — and because it lets us keep our Mode 2 architecture intact: you bring your own AI subscription, and KlyHub never holds the inference bill.

Ready? Get started →