Connection manual

Even without connecting, decisions and reports still reach matchwire. Connecting is an optional step. The only thing connecting adds is that you can enter and leave the same room from the AI you normally use—it does not mean AI will move anything forward on your behalf.

What connecting changes—and what it does not

When you connect an AI client you normally use, such as Claude Code, ChatGPT, or Codex, that AI can enter and leave the same room as matchwire. Building your profile, expressing intent, and making decisions all remain complete inside matchwire whether or not you connect.

Who can see what is set on the Privacy page. What a connected AI may do is set in two places only: the authorization screen your browser opens the first time you connect, and Settings → Connections, where you can disconnect at any time. There is no other gate—no extra confirmation and no per-action permission.

Instructions by client

• Claude — add matchwire as a custom connector on claude.ai or use the Claude Code CLI.
• ChatGPT and Codex — add matchwire to the shared MCP configuration over Streamable HTTP.

The first thing to ask after connecting is the same for every client. Each client page ends with the example.

For hiring-team staff

matchwire has separate candidate and hiring-team doors, and their connection URLs are different. The candidate URL ends in /mcp; the organization staff URL ends in /mcp/org. See Connect as a hiring team for the staff flow.

Where to find the connection URL

The URL depends on the environment hosting matchwire. After signing in, open Settings from the avatar menu. The Connections tab always contains the Use it from your usual AI card. Its Connection URL field shows the environment's URL ending in /mcp, with a copy button. The client instructions refer to this URL. The same card appears at the end of onboarding on the ready page.