Wedge for Claude

Give Claude a live connection to your startup

Your AI assistant doesn't know your startup. Every session starts with you re-pasting the same context — who the customer is, what you've already tested, what's actually blocking you — and by the time you're done you've spent more effort explaining the problem than thinking about it. Wedge fixes that. Your project state lives in Wedge, and the Wedge MCP server gives Claude a live, read-and-write connection to it, so every answer starts from what's true about your project today.

Start free Set it up

What this actually does

The same three questions every founder asks their assistant, answered with and without a connection to real project data.

“What should I work on this week?”

Without Wedge

Generic advice: talk to users, build an MVP, define your ICP. True of every startup, useful to none.

With Wedge connected

Reads your project's stage, progress, current bottleneck and open tasks, then answers against the specific thing blocking you right now.

“Is this idea validated enough to start building?”

Without Wedge

A confident guess based on the two paragraphs you pasted into the chat before running out of patience.

With Wedge connected

Pulls your Validation Score, the evidence behind it, your logged interviews and your problem sizing — and names the gaps you still haven't closed.

“Who exactly are we selling to?”

Without Wedge

A plausible persona invented on the spot, slightly different from the one it invented yesterday.

With Wedge connected

Reads the personas you defined and the interviews you logged, quotes what people actually said, and flags where segments contradict each other.

What is MCP?

MCP is the Model Context Protocol, an open standard created by Anthropic that lets AI applications connect to external tools and data sources through one common interface. Instead of every assistant inventing its own plugin format, an application exposes an MCP server once and any MCP-capable client can talk to it. Wedge runs one of those servers.

Because the standard is open, connecting Wedge to Claude is not a commitment to Claude. The same server URL works in Claude Code, Cursor, VS Code, Windsurf and anything else that speaks MCP, and your project state stays in Wedge either way. If you change assistant next month, you paste one URL into the new one and carry on.

Set it up

Every client uses the same server URL. Connecting runs an OAuth sign-in with your Wedge account, so the assistant only ever sees your own projects.

Your Wedge MCP server URL

https://sxqbykxrazwqsxsfbbtb.supabase.co/functions/v1/mcp

Claude (web and desktop)

Open Settings → Connectors and choose “Add custom connector”.
Name it Wedge and paste the server URL below.
Click Add, then complete the Wedge sign-in prompt.
Enable Wedge from the chat composer and ask it about your project.

Server URL

https://sxqbykxrazwqsxsfbbtb.supabase.co/functions/v1/mcp

Claude Code

Run the command below in a terminal.
Start Claude Code and run /mcp to confirm Wedge is connected.
Sign in from that menu when prompted.

Terminal command

claude mcp add --scope user --transport http wedge 'https://sxqbykxrazwqsxsfbbtb.supabase.co/functions/v1/mcp'

Cursor

Open Settings → MCP → “Add new global MCP server”.
Add the entry below to ~/.cursor/mcp.json.
Reload Cursor and approve the Wedge sign-in.

~/.cursor/mcp.json

{
  "mcpServers": {
    "wedge": {
      "url": "https://sxqbykxrazwqsxsfbbtb.supabase.co/functions/v1/mcp"
    }
  }
}

VS Code

Open the Command Palette and run “MCP: Add Server” → HTTP.
Paste the server URL, or add the entry below to .vscode/mcp.json.
Start the server from the palette and approve the Wedge sign-in.

.vscode/mcp.json

{
  "servers": {
    "wedge": {
      "type": "http",
      "url": "https://sxqbykxrazwqsxsfbbtb.supabase.co/functions/v1/mcp"
    }
  }
}

Using something else? Any client that supports remote MCP servers works — create a custom or remote server connection, paste the URL, and finish the sign-in.

What Claude can do with your Wedge project

Reading is the bulk of it. A smaller set of actions writes back, so a conclusion reached in chat lands in the workspace instead of scrolling away.

Read

Wedge project data Claude can read
Your projects	List every idea in your workspace with its stage, progress, current bottleneck and next action.
A single project in detail	Pull the full record for one idea — the idea statement, stage, and where it stands.
Validation score	The score, the evidence behind it, and the gaps still unproven before it's safe to build.
Customer personas	The segments you've defined, with the assumptions attached to each.
Customer interviews	Every logged discovery conversation — subject, segment, notes and signal strength.
Problem sizing	Scale, impact and how people solve the problem today, as you recorded it.
Waitlist signups	Demand collected from your validation landing pages.
Canvas notes	The raw working notes and cards on your canvas, not a summary of them.
Tasks	Open and completed work, optionally filtered by status.
Competitors	The landscape you mapped, with positioning and gaps.

Write

Changes Claude can make to your Wedge project
Add a canvas note	Drop a conclusion from your conversation straight onto the project canvas.
Log an interview	Record a customer conversation — subject, segment and notes — without leaving the chat.
Create a task	Turn a decision made in chat into a task on the project.

See it in action

Demo coming soon — a walkthrough of Claude reading a live Wedge project and writing a decision back to the canvas.

Frequently asked questions

What is MCP?

MCP is the Model Context Protocol, an open standard created by Anthropic that lets AI applications connect to external tools and data sources through one common interface. Wedge runs an MCP server, so any MCP-capable assistant can read and write your project state.

Does this work with tools other than Claude?

Yes. MCP is an open standard, not a Claude feature. The same Wedge server URL works in Claude, Claude Code, Cursor, VS Code, Windsurf and any other client that supports remote MCP servers.

Is my data sent to Anthropic?

Only what your assistant reads during a conversation, and only when you ask it to. Wedge does not push data anywhere on its own — your AI client requests specific project data over the connection, so it is handled under whichever assistant you chose to connect. Nothing is shared with any assistant you have not connected.

Can I use Wedge without the MCP server?

Yes. Wedge is a full workspace on its own — canvas, validation scoring, interviews, competitors, roadmap and tasks all work in the browser. The MCP server is an optional extra way to reach the same data.

What happens if I switch AI assistants?

Nothing breaks. Your project state lives in Wedge, not in a chat history. Remove the connector from the old client, paste the same server URL into the new one, and your assistant picks up exactly where the last one left off.

Do I need to sign in for the connection to work?

Yes. Connecting runs an OAuth sign-in with your Wedge account, and the assistant only ever sees the projects that account can already see.

Stop re-explaining your startup every session

Set up a Wedge project, connect it to Claude in about a minute, and let your assistant work from your real stage, evidence and bottleneck instead of a fresh guess each time.

Start free All integrations