MCP lets an AI assistant outside SMTasker work with information from your SMTasker account. Instead of copying details into a chat, you give the assistant a controlled connection so it can use the SMTasker tools available to it.

This walkthrough shows you where to turn access on, create a personal key, enter the connection details, and check whether it worked. Start with read-only access: your assistant can inspect supported data without permission to change it.

What can I use this for?

Think of MCP as a connection between two apps: your AI assistant asks for information, and SMTasker responds through the tools it makes available. MCP stands for Model Context Protocol. You do not need to learn the protocol to follow the setup.

• Ask questions with workspace context. Have the assistant summarize the SMTasker information it can actually retrieve, instead of relying on your description alone.
• Investigate an issue. Ask it to inspect relevant available data and explain what to check next.
• Request supported changes later. If you deliberately grant write access, the assistant can use supported tools that change your setup.

For example, you might ask, “What can you see in my SMTasker workspace, and what needs attention?” The answer depends on the available tools and your key’s permissions. Connecting MCP does not make every feature of SMTasker available automatically.

Do I need MCP, or should I use Skye?

Use Skye if you want to chat inside SMTasker. Open the purple Skye button and follow the Skye walkthrough. You do not need to set up an external MCP connection just to use Skye.

Use MCP if you want to work from a compatible external AI app. This guide connects that app to SMTasker.

Before you start

Have SMTasker open in one window and your external AI app in another. The AI app must support a remote MCP connection using Streamable HTTP and a Bearer authorization header. These are the connection options you will copy below—not code you need to write.

Client menus and supported authentication methods vary. This guide shows the real SMTasker screens; it does not assume that every AI app has the same connection form. If yours does not offer these options, check its connection documentation or ask support before creating a key.

Step 1: Find API & MCP Access

1. In the SMTasker sidebar, click Settings.
2. At the top of Settings, select API & MCP Access.
3. Find the MCP access switch. Turn it on when you are ready to connect your assistant.

What should you see? The page says Enabled and “Connections are allowed for valid API keys.” Enabling the switch alone does not connect an assistant: it will also need a valid personal key.

Step 2: Create a read-only personal key

A personal key is a private access credential. Treat it like a password. Give each assistant its own named key so you can stop one connection without affecting the others.

Scroll to API keys and click New API key.

1. Key name: enter a recognizable label, such as “My desktop assistant.” This label is not the secret key.
2. Expiration date: choose when access should end. Plan to replace the key in your assistant if you need access beyond that date.
3. Read data: leave this checked.
4. Make changes: leave this unchecked for your first connection.
5. Review the fields, then click Create key. Follow SMTasker’s instructions to copy the generated key and store it securely.

Step 3: Enter the connection in your AI app

Return to Connection details on the same SMTasker page. Keep it open while you add an MCP server or connection in your external AI app.

Transfer these details into your assistant’s connection form:

For the header value, type Bearer, then one space, then paste your generated SMTasker key. Replace the placeholder completely; do not include angle brackets.

If the AI app wants one complete header line instead of two separate fields, use this format:

Authorization: Bearer YOUR_SMTASKER_API_KEY

Copy header template copies the format, not your secret. Replace <personal-api-key> with the key you created. Save the connection, then use your client’s connect or test control.

Step 4: Ask a small first question

Once your AI app reports a connection, open a conversation where the SMTasker connection is enabled. Start with:

List the SMTasker tools available through this connection. Explain in simple terms what information you can read. Do not change anything.

What are you looking for? A response based on the connection’s actual tools, rather than a generic description of SMTasker. Then try:

Use the available read-only SMTasker tools to summarize my workspace. Tell me what needs attention and what I should check first. If you cannot retrieve something, say so. Do not make changes.

These are suggested first prompts, not screenshots of a tested external-client conversation. Your assistant’s response and approval controls depend on the client, the tools it can access, and your workspace.

If it asks you to paste all your workspace information manually, check that the connection is enabled for that conversation and that the client can list its tools.

Step 5: Check that SMTasker received the request

Return to Settings → API & MCP Access. Click Activity log at the top-right, then Refresh.

Use the time, key name, operation, HTTP status and outcome to match entries to your request. The log covers authenticated requests and executed batch actions, including reads, writes and rejections. An entry by itself does not mean the requested task succeeded—check its outcome and your assistant’s reply.

If nothing appears, refresh and check the URL, key and client setup. Unknown invalid credentials are recorded in the server security log rather than this view, so an empty list is not proof that no connection was attempted.

Allowing changes later

Only use a key with Make changes when you want that assistant to perform supported updates. Check the external app’s approval controls first: Skye’s Ask for permission setting does not control an external MCP assistant.

For a first write task, choose one specific target. Ask the assistant to explain the proposed change, review the exact account and automation, and use the client’s approval controls. Afterwards, open the relevant SMTasker screen and verify what was saved. Do not assume that connecting grants every product capability.

Stop access when you no longer need it

• Stop one assistant: find its named key in the API keys table and click Revoke. Other keys can remain active.
• Block all MCP connections: turn off Enable MCP access.
• If a key is exposed: revoke it and create a replacement for the intended client. Do not reuse the exposed key.

Removing access prevents future authorized use; it does not undo changes already completed. The key table also shows permissions, last use, expiry and status so you can review your connections.

If something does not work

• New API key is disabled: check the MCP access switch first.
• The AI app rejects the connection: verify support for remote Streamable HTTP with a Bearer header. Check that you used the SMTasker key, not an AI-provider key.
• Authentication fails: check the exact server URL, key status and expiry, and the space after Bearer.
• Reading works but changes fail: a read-only key cannot authorize writes. Only grant write access if you actually intend to allow changes.
• Requests are rate-limited: slow down the assistant’s requests and simultaneous tasks before retrying.
• Email or scheduling is unavailable: the current MCP panel explicitly lists these as unavailable; a new key does not add them.

The verified panel shows 120 calls per minute per key, 300 per customer, 30 changes per minute, and four concurrent calls per customer. Account sync has a separate limit. Check the panel for current values rather than relying on a screenshot indefinitely.

Still stuck? Contact SMTasker support with your AI app’s name, the step where you stopped, the error, and the approximate time. Include a screenshot with the secret removed—never send the API key itself.