Getting Started

This guide takes you from a fresh KlyHub account to a working MCP connection in five steps. The walkthrough uses Claude Desktop as the example client, but the steps are the same for any MCP-capable AI (see the MCP setup section for client-specific snippets).

You will need:

• A KlyHub account (sign up at app.klyhub.com).
• An AI client that speaks MCP — Claude Desktop, Cursor, ChatGPT (custom GPTs), or your own.
• About five minutes.

Step 1 — Create your tenant

When you sign up, KlyHub asks you to name your first workspace. The slug you pick becomes your tenant URL: <your-slug> will appear in the MCP endpoint as mcp.klyhub.com/v1/<your-slug>.

Pick something short and memorable. You can rename it later from Settings → Workspace, but the slug itself stays the same.

Step 2 — Run the intake

KlyHub does not ask you to fill out blank forms. Instead, the 4-phase AI-guided intake interviews you to populate the knowledge graph:

1. Foundation — who the company is, what it does, who it serves.
2. Context — market, competitors, positioning.
3. Exploration — current operations, motions, customers.
4. Optimization — gaps, priorities, what changed last month.

You do not have to finish the intake before connecting MCP — you can run it in chunks and your AI client will see whatever is already saved.

For the full methodology, see Intake methodology. For the data shape, see Ontology.

Step 3 — Authorize your AI client

KlyHub uses OAuth 2.1 for MCP clients. From inside your workspace:

1. Go to Settings → MCP clients.
2. Click Connect Claude Desktop (or your client of choice).
3. KlyHub opens a consent screen on accounts.klyhub.com. Approve the requested scopes.
4. Once approved, KlyHub shows you the client config snippet to paste into your AI client.

The token issued to your client is scoped to this tenant only — it cannot read or write any other workspace.

Step 4 — Paste the config into your client

Copy the snippet from Step 3 into your client's MCP config file. For Claude Desktop, that file lives at:

• macOS — ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows — %APPDATA%\Claude\claude_desktop_config.json

It should look like this:

{
  "mcpServers": {
    "klyhub": {
      "url": "https://mcp.klyhub.com/v1/<your-tenant-slug>"
    }
  }
}

Restart Claude Desktop. The klyhub server should appear in the tools panel.

For Cursor, ChatGPT, and custom clients, the exact file path and JSON shape differ slightly — see the MCP setup section.

Step 5 — Ask your AI a question

Try something concrete:

"What are our company's top three differentiators against [competitor], according to our knowledge base?"

Your AI client will call the klyhub.query MCP tool against your tenant, pull the relevant Market-layer entries, and answer with citations to the entries it used. If you haven't run the intake yet, it will tell you the layer is empty — that's your cue to head back to Step 2.

You're done

Ready to keep going? Open KlyHub → and run the intake, or dig into the Ontology reference to see what kinds of facts you can store.

If you hit something unexpected, check the API reference or open an issue on the project tracker.