TopolearnMany tutorials online still show Gumloop's old workflow canvas; knowing the current shape saves confusion.
Imagine a new assistant who can read your inbox, look things up in your CRM (customer database), write a summary and post it in Slack, and who learns your preferences over time. Gumloop lets you create that kind of assistant without programming.
In Gumloop such an assistant is called an agent: an AI model that is given a goal, a set of tools (actions it is allowed to take in your apps) and instructions, and then decides by itself which steps to take. You talk to it in plain language, and it works on the request until it is done or needs your input.
The main building blocks you will learn:
- Connectors: give an agent access to apps such as Gmail, Slack, Notion or Salesforce.
- Skills: reusable written playbooks for specific jobs.
- Brain: your company's documents, searchable by agents.
- Triggers: start an agent automatically, on a schedule, when an email arrives, or when another system calls it.
- Channels: where people use an agent, such as Slack, Teams, email or a web page.
- Credits: the unit Gumloop uses to measure usage.
A short history helps: Gumloop began as a workflow builder, where you dragged boxes (nodes) onto a canvas and connected them. By 2026 it had become agent-first. The workflow builder is still available under a "Workflows (Legacy)" heading, and older YouTube videos and blog posts mostly show it. This course teaches the current agent product, with one lesson on legacy workflows.
Every account also gets Gumball, a personal agent ready to use.
Remember
- Gumloop is now an AI agent platform; the workflow canvas is legacy
- An agent = model + instructions + tools, deciding its own steps
- Connectors, skills, Brain, triggers and channels are the building blocks
Watch out for
- Following a 2024 or early 2025 tutorial and looking for a canvas of nodes that new accounts are not centred on.