Model Context Protocol

Connect Claude and ChatGPT to Idea Kill Switch

Wire your favourite AI assistant to the Idea Kill Switch MCP server. Pull verdicts, validate new ideas and trigger market analysis straight from chat - no copy-paste, no custom plumbing.

MCP endpoint

https://api.ideakillswitch.com/v1/mcp

Authentication

OAuth 2.1 via Supabase

Setup time

About 5 minutes

Native MCP

Connect Idea Kill Switch to Claude Desktop or Claude.ai

Recommended path. Native MCP support, no developer toggle required.

Model Context Protocol docs

Open Claude settings
In Claude Desktop choose Settings -> Connectors. On claude.ai go to Settings -> Connectors. You need an active Claude account that supports custom connectors.
Add the Idea Kill Switch connector
Click "Add custom connector" and paste the MCP server URL below. Claude will read our /.well-known discovery document and configure the OAuth flow automatically.
MCP server URL
https://api.ideakillswitch.com/v1/mcp
Sign in and approve the requested scopes
Claude will redirect you to Idea Kill Switch. Sign in with your existing account and review the requested permissions on the consent screen - typically mcp:read for analysis access and mcp:write to create or update ideas. Approve to finish linking.
You can revoke the connector at any time from your account settings.
Start a conversation
Mention Idea Kill Switch in any chat and Claude will surface the tools it can call - for example "list my latest ideas" or "validate a new idea about ___". Tool calls execute against your account in real time.
Example prompt
List my Idea Kill Switch verdicts from the last 30 days and summarise the riskiest one.

How we keep your data safe

The Idea Kill Switch MCP server is a protected resource. Clients authenticate through OAuth 2.1 using Supabase as the authorization server and only receive the scopes you approve on the consent screen - typically mcp:read for read access and mcp:write for tool calls that modify your ideas.

Every tool call is scoped to your account, audited server-side and protected by the same Postgres row-level security as the dashboard. You can revoke any connector at any time from your profile.

Frequently asked questions

What is MCP and why does Idea Kill Switch support it?

Model Context Protocol (MCP) is an open standard that lets AI assistants call tools and read data from third-party services through a shared interface. Exposing Idea Kill Switch as an MCP server means Claude, ChatGPT and other compliant clients can pull your ideas, verdicts and market analysis directly into the chat without copy-paste.

Do I need a paid Idea Kill Switch plan to use the MCP connector?

You need an Idea Kill Switch account in good standing. Read-only tool calls work on any active plan. Tool calls that consume analysis credits - for example running a fresh validation - follow the same billing as the web app.

How does authentication work?

We implement OAuth 2.1 via Supabase as the authorization server, with the MCP server published at api.ideakillswitch.com/v1/mcp as the protected resource. Clients discover the auth server through our /.well-known/oauth-protected-resource document. Tokens are issued per client and can be revoked from your account at any time.

Which scopes does the connector ask for?

Most clients request mcp:read to read your ideas and analysis reports and mcp:write to create or update ideas on your behalf. You approve them explicitly on the consent screen before the connector is linked.

My client is on the older MCP spec - will it still work?

Yes. We publish both the new RFC 9728 protected-resource metadata and the legacy RFC 8414 authorization-server document at the well-known URLs, so clients on the 2025-06-18 and 2025-11-25 specs both discover the right endpoints.

Ready to plug Idea Kill Switch into your AI workflow?

Create an account, generate your first verdict, then point Claude or ChatGPT at the connector. The whole loop takes about 10 minutes.