Use your Promta prompts in your AI tools

Connect Claude Code

1. 1
   In Promta, open Settings → Connections and turn on Allow AI tools to use your prompts.
2. 2
   Click Copy Command.
3. 3
   Paste it into Terminal and press Return. The command carries your own port and token, so don’t retype it by hand.
4. 4
   Check it worked — run /mcp inside Claude Code, or claude mcp list in Terminal. You should see promta connected.

Now try it. Ask: “What prompts do I have saved about writing?” or “Use my code review prompt on this file.”

Connect any other MCP tool

Most MCP tools are set up with a config file rather than a command. Turn on the connection in Settings → Connections, click Copy Configuration, and paste the result into your tool’s MCP config. You’ll get something like this, with your own port and token filled in:

{
  "mcpServers": {
    "promta": {
      "type": "http",
      "url": "http://127.0.0.1:PORT/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

The three values are always the same — the address, the Authorization header, and the name promta. What changes is what each tool calls them, which is the most common reason a config that looks right doesn’t connect.

{
  "mcpServers": {
    "promta": {
      "url": "http://127.0.0.1:PORT/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

{
  "servers": {
    "promta": {
      "type": "http",
      "url": "http://127.0.0.1:PORT/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

{
  "mcpServers": {
    "promta": {
      "serverUrl": "http://127.0.0.1:PORT/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

{
  "mcpServers": {
    "promta": {
      "httpUrl": "http://127.0.0.1:PORT/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

What’s been tested: Claude Code and the official MCP Inspector, directly against Promta. The configurations above come from each tool’s own documentation, linked beside it — if one has changed its format, the docs are right and this page is stale, so trust them and let us know.

If your tool takes the address and header in a settings screen instead of a file, use Copy address and Copy token from the same pane.

What can’t connect yet

Tools that only speak stdio. Claude Desktop is the common one — its config accepts a launch command, not an address, so it can’t reach Promta over HTTP. Support for it is coming.

Cloud assistants that run on someone else’s server. Anything that connects to MCP servers from the cloud needs a public address it can dial. Promta deliberately listens only on your own machine, so there is nothing for a remote service to reach — which is the same property that keeps your prompts private. Tools that run on your Mac are the ones that work.

What a connected tool can do

• Search your prompts by text, tag, or AI model — free.
• List your favorites and see which tags you use — free.
• Fetch the full text of any prompt, including the variables it needs — free.
• Save a new prompt to your library — free, up to your plan's prompt limit.

If a prompt uses variables like {{language}}, the tool asks you for the values — or offers your choices, if you set options — before using it, the same way Promta does.

It can also save new prompts — that’s free too. Ask your tool to save a prompt and it appears in your library on every device. It counts against your plan’s prompt limit just like a prompt you save by hand (25 on the free plan; unlimited on Pro).

What it can never do is change what’s already there. No connected tool can edit or delete an existing prompt, so your library can’t be rewritten behind your back — by a mistake or by instructions hidden in something you saved.

Is this private?

Promta listens only on your own machine’s loopback address — the connection isn’t reachable from any network, including your local Wi-Fi. Every request needs a token that Promta generates and only you can see, and you can regenerate it at any time to disconnect every tool at once.

The connection is off until you switch it on, so if you never use this, nothing is ever opened.

If something isn’t working

“Promta isn’t running”

Open Promta. The connection only exists while the app is running. To avoid this, turn on Launch at Login in Promta’s settings.

Your tool says the server failed to connect

Check that “Allow AI tools to use your prompts” is still on in Settings → Connections. If you regenerated your token, the old details stopped working — copy them again. In Claude Code, run claude mcp remove promta before adding it back.

It connects but sees no prompts

Confirm you have prompts saved in Promta. If your library isn’t empty, toggle the connection off and on to restart it.

Your tool has nowhere to put a header

Then it can’t connect — that’s how Promta authenticates. Some tools only allow headers on remote servers; look for an “advanced” or “custom headers” option before giving up.

Still stuck? Email [email protected].