PublishPort

npx add-mcp 'https://api.publishport.app/mcp'

Send your AI one line — it installs the PublishPort Skill and hands you an authorization link; one click and it can publish through your machine. MCP setup for Cursor, WorkBuddy, Cline, Claude, Codex and others is covered too.

Send your AI this one line — Claude Code, Codex, Cursor, OpenClaw, Gemini CLI or any other AI that can run commands:

Read https://www.publishport.app/skill.md and follow it to set up PublishPort for me.

The same line sits on the Connect AI page in the app, with a Copy button.

What happens next

1. Your AI reads the instructions and installs the PublishPort Skill into its own skills folder, so future sessions already know how to publish.
2. It runs publishport login and gives you an authorization link and a confirmation code.
3. You open the link, sign in (or create an account on the spot), check that the code on the page matches the one your AI showed you, and click Authorize.
4. Your AI picks up the credential by itself, checks which of your devices are online and which platforms are signed in, and tells you what’s ready and what’s left.

The authorization page shows which device and which region the request came from. If you didn’t ask your AI to do this, click Deny. Authorizing lets that AI run commands through PublishPort on the computer where your desktop app is signed in; every AI you’ve authorized is listed under AI access on your account page, where you can revoke it at any time.

It works the same when your AI runs in the cloud: commands travel through the official relay and execute on your machine, wherever the AI itself lives. Just keep the desktop app running.

Use MCP instead

If your AI only speaks MCP and can’t run commands (some chat clients), or you simply prefer MCP, you can connect it by hand.

Open the Connect AI page in the app and expand Use MCP instead:

1. Confirm the status reads Local executor connected. (If not, go back to Install and check the app is running in the background.)
2. Click Generate address to get an MCP URL that looks like https://…/mcp, plus a ready-to-copy JSON config.

This URL is the key. Any AI that has it can drive your computer, so don’t paste it into public chats, screenshots, or code repositories.

One-click JSON config (Cursor / WorkBuddy / Cline, etc.)

Most GUI clients accept the same standard mcpServers JSON. The Connect AI page already fills in your address — hit Copy to grab the config, then paste it into your client’s MCP config file:

{
  "mcpServers": {
    "publishport": {
      "url": "https://…/mcp"
    }
  }
}

Replace https://…/mcp with your own access address (the one you copy from the app is already complete — no hand-editing needed).

Where to paste it, per client:

• Cursor: Settings → MCP → Add new MCP server, or edit .cursor/mcp.json (project) / ~/.cursor/mcp.json (global) directly and merge in the mcpServers block above.
• WorkBuddy: paste the same JSON into its MCP / tool-server settings.
• Cline (VS Code extension): open the Cline panel’s MCP Servers → Configure → Edit Configuration and add the block to cline_mcp_settings.json.
• Any other GUI client: the entry point is named differently (“Add MCP server / tool server”), but nearly all accept this mcpServers + url JSON.

Two common variants

Some clients use different field names — if a straight copy won’t connect, tweak it as below:

• Windsurf: rename url to serverUrl (file lives at ~/.codeium/windsurf/mcp_config.json):

  {
    "mcpServers": {
      "publishport": {
        "serverUrl": "https://…/mcp"
      }
    }
  }
  

• VS Code (native MCP, not Cline): the top-level key is servers (not mcpServers), and you must set "type": "http" — otherwise VS Code treats it as a local stdio server and fails (file lives at .vscode/mcp.json):

  {
    "servers": {
      "publishport": {
        "type": "http",
        "url": "https://…/mcp"
      }
    }
  }
  

Command-line tools (Claude Code / Codex)

CLI tools don’t need hand-edited JSON — the Connect AI page gives you a command to copy straight into your terminal:

• Claude Code:

  claude mcp add --transport http publishport "<your-address>"
  

• Codex:

  codex mcp add publishport --url "<your-address>"
  

Connect any other MCP client

Claude desktop, Manus, and any AI that speaks MCP can use the universal address: copy it from the Connect AI page and add it wherever that client lets you “add an MCP / tool server (HTTP transport).” The entry point is named differently across apps, but it’s always “add an MCP server,” then paste the address.

Already have an address? Give it to the CLI

An address you generated also works with the official CLI, skipping the browser step:

npx publishport login "<your-address>"

The credential lives in ~/.publishport/config.json (mode 600). It’s the same key as the address — don’t share it.

Confirm it’s connected

Once connected, have your AI take a look at what’s available — try asking:

Which platforms can you publish to right now?

Your AI will list the platforms that are logged in and available on your machine. See the list, and the channel is open — time for your first AI publish.