What is an automation action?

An action is a step that runs after a trigger fires. Taskade ships hundreds of them, and the ones worth knowing are not the ones that copy a field from one app to another — they are the steps that read the web, reshape the data, and hand a decision to an agent that remembers the last time it made one. The 16 workflows below are grouped by capability rather than by app. Each puts a family of steps on the page: the agent lifecycle, the research steps, the data transforms, and the project and task operations that write the result back where your team works. An action is one step in a running automation. A trigger starts the run and hands over an event; each action then does one thing with it — search the web, convert a PDF to text, sum a column, ask an agent, create a task — and passes its output to the next step. A workflow is just an ordered chain of them. Actions come from two places. Connected apps contribute the steps their API exposes, which is what the category index covers app by app. Built-in Taskade actions contribute the rest, and need no account to connect: Taskade AI for agent calls, Media for research, Utilities for data work, and the workspace itself for projects and tasks.

How is a Taskade action different from a data-mover step?

Most automation tools move data between apps. The step in the middle is stateless: it runs, it returns, it forgets. That is a genuine capability, and Taskade ships it too. The difference is what else you can put in that slot. A Taskade action can address a named agent that keeps its own knowledge and its own memory between runs. "Ask Agent" is not "ask the model" — the agent that reads today's contract has read every contract before it, and it carries the red lines you gave it. Run the same step tomorrow and it starts from what it learned, not from an empty prompt. The multi-agent workflows go further and hand one job to a team of them.

What can the built-in steps do?

Research — search the web, scrape a page, summarize a site, transcribe a video, turn a PDF into plain text. Transform — parse CSV into JSON, pull a value out of a JSON payload, sum or average a column, match a regex, format a date, slugify a title, resolve a lookup table, generate a QR code. Reason — ask an agent, run a saved agent command, ask an agent team, respond in an agent chat. Write back — create a project from a template, tag it, set its date range, add tasks, assign them, set due dates, move them between sections, update custom fields. Chain them and a single run can go from a raw webhook to an owned task with a due date, with the reasoning attached.

Who are these workflows for?

Anyone whose automations currently stop at "copy this field over there". Operations and support teams get the triage and queue steps, finance and legal get the document and totals steps, marketing and research get the web and video steps, and every team gets the project and task steps that make the output someone's job rather than a notification. Browse the rest of the library by category, see the end-to-end workflows and repeatable processes they slot into, or start from the automation hub.

How do you build one?

Open Automations in your workspace, add a trigger, then add the actions in the order you want them to run. Every step's output is available to the steps after it, so an agent can read what the scrape returned and a task can carry what the agent decided. Or describe the outcome in plain language and let Taskade wire the chain for you — start from a prompt and edit the steps afterwards.