The model chooses, inside limits
A step can hand a model a conversation and a set of tools, and let it pick which to use. It cannot reach a tool its code did not offer it.
AI agent builder
Build an AI agent by describing it
Describe the agent’s job in plain words and AI builds it. Code does the steps you already know, a model handles the ones that need judgment, and you can see which is which before it runs. The same input takes the same path, and the agent can act through any of 1,400+ tools.
From a job description to a running agent
Three steps, whether the agent answers questions, triages an inbox or writes a report.
- DescribeSay what the agent does
What starts it, what it may use, what a good result looks like, and where a person should decide.
- ReviewSee where the model decides
AI writes the agent once, as code. The map shows each step, and which of them ask a model.
- PublishRun it on a trigger
On a schedule, on a webhook from another app, or on each message in Slack or Telegram.
Agents to start from
Real briefs for the same kind of job. Pick one and Whenever writes it against your accounts; the code is yours to change from there.
Track keyword rankings in a sheet each weekChecks the search position of every keyword on your list, appends the history to a sheet and reports the moves.
Build this workflowAsk questions about a Google Sheet in SlackAnswers a plain-English question about a worksheet when someone on your allowed list mentions the bot in Slack, using only the sheet's rows and naming the figures it used.Build this workflowSend yourself a morning digest on TelegramDelivers one Telegram message each morning with your unread email, today's meetings and the tasks due.
Build this workflowGet a conversion audit of any landing pageFetches a submitted page and emails back a review of its headline, call to action and friction points.
Build this workflowRun a coaching chat that remembers sessionsReplies to coaching messages on Telegram using each person's profile and history from a sheet, and emails everyone a check-in every Monday. After publishing, you pick a secret token in Whenever and pass it with the webhook address to Telegram's setWebhook; one bot serves one workflow.Build this workflowPost a new video to your channels every dayExpands the next idea in a sheet into a short video, publishes it to your accounts and marks the row done.
Build this workflow
Build this workflowAsk questions about a Google Sheet in SlackAnswers a plain-English question about a worksheet when someone on your allowed list mentions the bot in Slack, using only the sheet's rows and naming the figures it used.Build this workflowSend yourself a morning digest on TelegramDelivers one Telegram message each morning with your unread email, today's meetings and the tasks due.Build this workflowGet a conversion audit of any landing pageFetches a submitted page and emails back a review of its headline, call to action and friction points.Build this workflowRun a coaching chat that remembers sessionsReplies to coaching messages on Telegram using each person's profile and history from a sheet, and emails everyone a check-in every Monday. After publishing, you pick a secret token in Whenever and pass it with the webhook address to Telegram's setWebhook; one bot serves one workflow.Build this workflowPost a new video to your channels every dayExpands the next idea in a sheet into a short video, publishes it to your accounts and marks the row done.Build this workflowWhat makes it an agent you can trust
Judgment where it helps; code everywhere else.
The rest is code
Looking up the order, writing the row, sending the email: plain code that runs the same way every time and spends no tokens.
You pay for the judgment only
AI building is free within weekly usage limits. During runs, model tokens are spent only in the steps that ask a model, and draw on the separate runtime AI allowance every plan includes before they are billed on top.
An agent on Whenever, or an open-ended agent?
Build it on Whenever when
- The job has known steps and one or two that need judgment.
- A wrong path on run fifty-seven would cost you a customer.
- You want to read what the agent does and keep it as a file.
Use an open-ended agent when
- Working out the steps is the job itself, and they change every time.
- A different answer on every run is acceptable, or the point.
- Nobody needs to review the path before it is taken.
A few useful answers
What is an AI agent builder?
A tool for making software that takes a job, decides some of the steps itself and acts in your apps. On Whenever you describe the job, AI writes the agent once, and a model is asked only where the job needs judgment.
Can I build an AI agent without coding?
Yes. Describe what the agent does in plain words, test it against real inputs, and publish it. What Whenever writes is code, which you can read or hand to a developer, but you never have to write it.
Does the agent call a model on every run?
Only in the steps that need one. A chat agent replies with a model on each message; a report agent may ask a model once to summarise and do everything else in code.
How is an AI agent different from an AI workflow?
An agent works out its next step as it goes; a workflow follows steps decided in advance. Most business jobs are a workflow with one or two judgment calls, which is what Whenever builds.
Sources & further reading
Sources reviewed September 30, 2026.
Describe the agent you need
Say what the job is and where it needs judgment. Whenever builds it and runs it the same way every time.