Most AI coding assistants and chat tools know nothing about your manufacturing process. To work with your live Threaded data—work instructions, the Map, parts and tools, issues, and more—you need a secure connection that runs as you, with the same permissions you have in the app.

Threaded MCP is that connection. It is a hosted server at https://mcp.threadedmfg.com that speaks the Model Context Protocol (MCP). After you sign in with your Threaded account, supported clients can search your organization, read process data, and run approved Threaded workflows on your behalf.

#Key Concepts

Threaded MCP — Threaded’s hosted MCP service. It exposes your organization’s data and workflows to external AI clients through a standard protocol, always acting as the signed-in user.

Hosted MCP server — The remote endpoint at https://mcp.threadedmfg.com/mcp (production). Clients connect over HTTP; you do not install or run a local server for production use.

OAuth sign-in — The first time a client connects, it opens a browser window so you can sign in with your Threaded account. The client keeps a secure session until you disconnect or it expires.

#Where You Can Connect

You can use Threaded MCP from three places:

Inside Threaded — the AI Assistant sidebar can use the hosted MCP tool catalog during chat.
Cursor — add the server in MCP settings or a project config file.
Claude Desktop or Claude Code — add the server as a remote HTTP MCP connection.

Each path uses the same hosted server and OAuth sign-in. Pick the one that matches how you work.

#Connecting in Threaded

The fastest way to use hosted MCP is inside Threaded itself.

#Open the AI Assistant

Click the AI Assistant icon in the top navigation bar. The chat sidebar opens from anywhere in the app.

#Connect Threaded MCP

Look for the Threaded MCP card below the chat header. If it is collapsed, click the cloud icon next to the model selector to expand it.
Click Connect when prompted. Threaded opens a sign-in window if you are not already authenticated for MCP.
When the status shows Connected, the assistant can use the hosted MCP tool catalog in that conversation.

If your organization supports session bootstrap, Threaded may connect automatically using your existing app login. If that fails, use Connect to complete OAuth manually.

#Disconnect

Expand the Threaded MCP card and click Disconnect when you want to end the session.

For an overview of how the in-app assistant works, see The AI Assistant.

#Connecting from Cursor

Cursor supports remote MCP servers over HTTP. You add Threaded once; Cursor handles OAuth the first time you use it.

#Cursor Settings

Open Cursor Settings → MCP (or Features → MCP, depending on your Cursor version).
Add a new MCP server.
Set the server URL to https://mcp.threadedmfg.com/mcp.
Save the configuration. When Cursor first talks to the server, it opens a browser window so you can sign in with your Threaded account.
After sign-in, confirm the server shows as connected in MCP settings. You can inspect MCP Logs in Cursor if something fails.

#Config File

You can also define the server in a JSON config file:

Project-level: .cursor/mcp.json in your repository root
User-level: ~/.cursor/mcp.json on your machine

Example for production:

{
  "mcpServers": {
    "threaded": {
      "url": "https://mcp.threadedmfg.com/mcp"
    }
  }
}

Restart Cursor or reload MCP after editing the file. The first tool call triggers OAuth; complete sign-in in the browser when prompted.

#Connecting from Claude

Claude supports remote MCP servers over HTTP with OAuth discovery. The production server URL is always https://mcp.threadedmfg.com/mcp.

#Claude Desktop

Open Settings → Connectors (or the MCP section in your version).
Add a custom remote server with the URL above.
When Claude connects, complete the Threaded sign-in flow in your browser.
Verify the connection is active before asking Claude to query Threaded data.

#Claude Code

Use the claude mcp add command with HTTP transport, or follow Claude’s current remote MCP setup flow for the same URL. Complete Threaded sign-in when prompted.

Exact menu names vary by Claude version; the server URL does not.

#After You Connect

Once connected, MCP clients can use Threaded on your behalf within your organization permissions. Typical capabilities include:

Search and read — list and fetch work instructions, Map nodes, parts, tools, issues, organization reports, and related objects.
Run Threaded CLI workflows — structured commands for tasks that would normally use the Threaded CLI, executed in a bounded sandbox as your user.
Check connection status — confirm which environment you are connected to and whether authentication is valid.

Destructive or write actions require appropriate OAuth scopes and your approval in the client where applicable. The in-app AI Assistant still uses its human-in-the-loop Execute flow for changes it proposes in chat.

For prompt ideas once you are connected, see AI Assistant Use Cases.

#Server URLs

Most customers use the production server:

Environment	MCP server URL
Production	`https://mcp.threadedmfg.com/mcp`

Threaded staff and preview environments may use https://mcp.threadedmfg.dev/mcp instead. Use the URL that matches the Threaded app you sign into (app.threadedmfg.com for production).

Visit https://mcp.threadedmfg.com in a browser for a short overview of the hosted service and its public endpoints.

#Troubleshooting

Connection failed or expired — disconnect and connect again from the Threaded MCP card, or remove and re-add the server in Cursor or Claude.
Wrong organization — MCP runs as the Threaded user you signed in with. Sign out and reconnect if you need a different account.
Client does not prompt for login — ensure the URL ends with /mcp and that your client supports remote HTTP MCP with OAuth.
Still stuck — contact your Threaded administrator or Threaded support with the client name (Cursor, Claude, or in-app) and the time the failure occurred.

#Why Connect Threaded MCP?

Connecting Threaded MCP brings your process data into the tools you already use for analysis, scripting, and automation—without exporting spreadsheets or copying context by hand. Your credentials stay scoped to your user, you can disconnect at any time, and the same hosted server powers both external AI clients and the in-app AI Assistant’s expanded tool catalog.