← Back 🔌

MCP Connections

Expose your business model as an MCP server so Claude Code, Claude, and Cursor can read it — and write deliverables back.

What it is

Your business model exposed as an MCP server, connecting out to Claude Code, Claude, and Cursor that read it and write back

LEANSpark exposes your business model as a Model Context Protocol (MCP) server. That means AI tools like Claude Code, Claude, and Cursor can read your canvas, segments, interviews, offer, competitive landscape, and voice profile — and write deliverables straight back into your active experiment. The work happens in the tool you already use; the validated context comes from LEANSpark.

Where it lives

Settings → MCP (/settings/mcp).

The MCP settings page with the server URL, the claude mcp add command, and the connected-tools list

How to use it

Connect a tool — copy the server URL and add it to Claude Code (claude mcp add --transport http leanspark <url>), Claude (Settings → Connectors → Add custom connector), Cursor (Settings → MCP → Add server), or any MCP client that supports remote servers with OAuth. During authorization you choose which idea the tool can access.
See connected tools — each shows the client, which idea it’s connected to, and when it was last used.
Revoke access — disconnect any tool; it has to reconnect to regain access.

What a connected tool can reach: it reads your business model, voice profile, segments, interviews, offer, competitive landscape, and product guide, and writes deliverables back into your active experiment. Every connection is scoped to a single idea and to your account only — nothing leaves the boundary you set.