Skip to main content

Build an AI automation in Claude with the qbash MCP

Alex MacLeod

This video connects Claude to the official qbash MCP server and builds a scheduled Task from one chat. The example is a small coffee roaster whose inventory lives in a Google Sheet. Claude reads the sheet, works out a four-week reorder grouped by supplier, turns that into a Task, posts the list to Slack and schedules it for 9am on weekdays. The Task keeps running when Claude is closed.

What's covered

  • Adding the official qbash MCP server to Claude as a custom connector
  • Connecting Google Drive from the chat and asking what's running low
  • Turning the reorder list into a Task
  • Connecting Slack and posting the list to a channel
  • Scheduling the Task for 9am on weekdays

Add the server to Claude

The official qbash MCP server works in Claude, ChatGPT, Codex, Cursor and other MCP clients with one URL, https://mcp.qbash.com/mcp. From a chat you can build Python Tasks, run them, read their logs, schedule them, use the apps you've connected and share memories across your AI tools.

In Claude, Alex opens Connectors, adds a custom connector called qbash and pastes the URL. He signs in with his qbash account and approves Claude's access. If you don't have an account yet, you can sign up from the same screen and you land back in Claude afterwards.

Read the inventory sheet

The sheet holds what's on hand, what sold last week and who supplies each item. Right now someone checks it by hand every morning to work out what to reorder. Alex asks Claude to connect his Google Drive, and qbash returns a link to sign in to Google. Back in the chat, he asks what's running low. Claude reads the sheet and lists how many weeks of stock each item has left and who supplies it. He then asks how much to reorder for four weeks of stock, grouped by supplier.

Turn the reorder list into a Task

That list is what he wants every morning, so he asks Claude to turn it into a Task. A card in the chat shows qbash writing the Task and testing it, and a run card follows along when he runs it. The output is the same list, now coming from the Task.

A Task is an AI cloud function written in Python that qbash hosts on secure cloud infrastructure. Opening it in qbash shows the Python qbash wrote and the log from the run Alex just did. There is a full editor there too, though everything in the video happens from Claude over MCP.

Post to Slack and schedule it

Next he asks Claude to post the list to a demo channel in Slack. Slack isn't connected yet, so he signs in from the link qbash returns. qbash updates the Task and runs it again, and the list shows up in the channel.

Last, he asks Claude to run it every weekday at 9am. The schedule runs on qbash, so the Task goes out every morning with Claude closed and the laptop off. Changing it later is another request in the same chat.

Official MCP server has the server URL and the setup steps for each client.

Open qbash →

Video transcript

So this is the official qbash MCP server. You can add it to claude, chat, GPT, codeex, cursor, whatever you use uh with one URL. And from your chat, you can build Python tasks. You can run them, look at their logs, schedule them to run. You can use apps you've connected and share memory across your agents.

All kinds of stuff. I'm going to show you what that looks like in Claude. going from a quick question to a task that runs every morning. So, first I copy the server address from here in claude. I go to connectors.

I add a custom connector called qbash and paste it in. Now, I already have a Cube Bash account, so I just sign in. If you don't, you can sign up right here, and you'll land back in Claude afterwards. I approve Claude's access and qbash is connected. In a new chat, I ask what qbash can do.

It works with the apps I connect, builds tasks that run on their own, and keeps memory. So, here's my scenario today. We're a small coffee roaster, and our inventory lives in this Google sheet. Uh, what's on hand, what sold last week, and who supplies each item? Right now, someone checks it by hand every morning to work out what to reorder.

I'd like that done for me. So, I ask Claude to connect my Google Drive. It gives me a link and I sign into Google. Back in Claude, I ask what's running low. It reads the sheet and shows how many weeks each item has left and who supplies it.

I want something I can act on. So, I ask how much to reorder for 4 weeks of stock grouped by supplier. That's what I want to run every morning. So, I ask Claude to turn it into a task. The card shows qbash writing it and testing it.

Then I run it and the run card follows along. Now, same list, but now it comes from a task. Let me open it in qbash. You don't have to look at this, but it's here if you want to review it. This is real Python that qbash wrote and the log from the run I just did.

There's a full editor in here too, but you can do all this from Claude over MCP. And that's how we like to do it, too. Back in Claude, I ask for one more change. Post the list to our demo channel in Slack. The Slack isn't connected yet, so I sign into Slack from the link.

qbash updates the task and runs it again. And here it is in Slack. Finally, I ask Claude to run it every weekday at 9:00 a.m. Now it runs on its own every morning, even when Claude is closed, even when my laptop is off. So, from one chat, I connected Drive and Slack, worked out the reorder list, and turned it into a scheduled task that I can change just by asking.

The official MCP servers address and the setup for each app are here on qbash.com under docs, official MCP server. Check it out.

Stay in the loop

Product updates, tutorials, and AI insights. No spam.