What you can do

This is what you can do with Toolscaled, in the order that you usually do it. How Toolscaled works tells the same story in five steps.

Connect tools

Use built-in tools, add tools from an OpenAPI document, or attach another MCP server. Group them into toolkits and add playbooks that explain how to use them. Store the credentials they need under Connections.

Explore a task

Ask the agent to do the work with a toolkit. Toolscaled records each call, its result and the data it used. The example prompts show how to ask.

Make a workflow

When the approach works, ask the agent to turn it into a workflow. You check the description, the inputs and the steps, then publish, and from then on the task no longer depends on you to repeat it. Steps that need a model to decide keep a model call; the other steps run as plain tool calls.

Run it again

Run the workflow from the chat, from the website, or as a tool the agent calls. A step can run once for each item in a list.

Automate it

Set when the workflow runs by itself: on a schedule, on a new Gmail or Outlook email, from a webhook or a form, or after another workflow. Nothing runs until you turn the automation on.

Answer action items

Decisions, missing credentials and failed runs appear under Action items on the website and in the console. Troubleshooting says what to do for each kind.