Automate routine work

When something happens, {{brand}} runs the steps you chose, as you, with your rights.

In short

An automation is When, Do and Then. When: a schedule, a webhook, something that happens, such as a check that starts failing, or a figure crossing a line you set. Do: one to eight steps - any action you may run, an AI step that writes or decides, a condition or a wait. Then: you hear about it when a run fails, or after every run.

Before you start

• Who: anyone signed in. An automation always runs as you, and each step checks your rights again right before it runs.
• Plan: Free keeps 3 automations and 100 runs a month; paid plans keep more. AI steps spend your AI allowance.
• First: nothing. Ready templates fill in most of it.

Steps

1. Open Products › Automations and press "New automation".
2. Templates opens. Pick a ready one, describe what you want and press "Build it with AI", or press "Start blank".
3. Check When, Do and Then. A step that sends or publishes needs you to tick "I confirm this step may send". Until you do, the automation saves switched off, and a line beside Save takes you to that step.
4. Press "Save". From then on, an open automation saves each change by itself. To try it now, press "Run now" and open the run under Runs.

Result: the automation runs on its own from now on, and every run shows each step, what it was given and what it did.

Good to know

• An automation never deletes anything.
• Values from earlier steps show by a plain name, such as "What step 1 wrote". Press "Use a value" inside a field to bring one in. A board or a chat is picked by name, and any other record is found by searching its name.
• "Remove" takes an automation away at once. Press Undo in the message that appears and it comes back as it stood. A webhook automation then asks for a new address.
• A new automation you have not saved stays for this session. If you leave it, Templates offers to continue it.
• After a save, a schedule or a watch is checked the way Test checks it, and what would stop a run shows beside its step.
• If you lose a right a step needs, the run stops and the automation switches itself off. You are told why.
• "Ask me first" makes each run wait for your yes before a step that sends.
• A run that failed can be run again from its page.
• Outside apps such as Zapier, Make or n8n can start an automation through its webhook address. The address shows once, right after you save, with a Copy button. "Get a new address" replaces it, and the old one stops working.
• Automations keep working when AI is off. An automation with no AI step goes on running and can still be edited. One with a write, decide or agent step is paused with that reason, never deleted, and comes back on by itself when AI is on again. Steps that need AI cannot be added while AI is off.
• A watch starts an automation when a figure crosses a line you set - for example overdue cards on your team above 5. Pick "When a figure crosses a line", the figure, the whole workspace or one of your teams, above or below, and the line. It is checked every 15 minutes against this month's value, starts once when the figure crosses the line and again only after it has come back, and costs no AI.
• Steps can branch: a condition can send the run to another step, and any step can say what happens if it fails - stop, carry on, or go to a later step. Runs only move forward, so they never loop.
• An action can repeat for each entry of a list, and a "Keep a value" step keeps a value for later runs. Repeating and where the run goes next are under "More options" on a step.
• An AI agent step works toward a goal with only the actions you list for it. It never deletes and sends only where you ticked it.
• Flow shows the steps as boxes and arrows. Test shows what each step would do without running anything. Versions keeps your last 20 saves.
• In an organization, its administrators can see every automation, with the teams of its owner, and switch one off. Team shows one team's automations, State the ones whose last run failed. Only you can change yours or switch it back on.
• When AI cannot build it - AI is not set up, or the provider did not answer - the ready ones closest to what you described are offered right there.
• If someone else saved the automation while you were editing it, you choose: Reload takes their version, Keep mine saves yours over it.
• A record an automation made shows that automation in its Related panel and on its Map, so you can see where it came from.
• AI steps send only what they need: long values are shortened and an agent step remembers every move it made in the run.
• Runs can be narrowed to one status and the last day, 7, 30 or 90 days. Versions shows what changed since the save before each one.
• Test can ask AI too: write and decide steps then answer as in a real run, so you see what later steps would get. It spends AI tokens and never runs an agent.
• A webhook automation can take only signed calls. Switch "Take only signed calls" on, copy the secret, and sign each call with the X-{{brand}}-Timestamp and X-{{brand}}-Signature headers. Unsigned calls are refused.
• Your organization's administrators can set a monthly AI limit for a module. When it is reached, AI answers nothing more there until the 1st of next month.
• Each automation with a write, decide or agent step has its own monthly AI limit, set under "AI limit for this automation" in the editor. The editor shows what this month used and, from the last 30 days of runs, about what a month costs - and warns when that pace passes the limit.
• An automation can also start when a fixed bug is verified, or when a card that waited for approval is approved or turned down - the decision is its status value, so a condition step can take either way.