Skip to content

Your First Workflow

A workflow on AgenFleet is a cron job: a scheduled task that runs your agent automatically and delivers the output somewhere useful. This guide walks you through creating one end-to-end.

We’ll build a daily news briefing — an agent that runs every morning, researches top stories in your industry, and sends a summary to your Telegram.


Make sure you’ve completed the Quickstart and have:

  • At least one deployed agent with web_search enabled
  • A Telegram account linked under Settings → Notifications

  1. Open your agent

    From the Fleet dashboard, click your agent’s name to open its detail view.

  2. Go to the Cron tab

    Select the Cron tab, then click New Job.

  3. Write the prompt

    The prompt is what the agent receives as its task when the job fires. Be specific:

    Search for the top 3 news stories in the AI industry from the last 24 hours.
    For each story, write:
    - A one-sentence headline
    - A 2-3 sentence summary
    - Why it matters for businesses adopting AI
    Keep the total output under 400 words. Format it cleanly for a Telegram message.
  4. Set the schedule

    Use a cron expression or the visual scheduler. For every morning at 8 AM:

    0 8 * * *

    Common schedules:

    ScheduleExpression
    Every day at 8 AM0 8 * * *
    Every Monday at 9 AM0 9 * * 1
    Every 6 hours0 */6 * * *
    Every weekday at noon0 12 * * 1-5
  5. Configure delivery

    Under Delivery, select Telegram and choose your connected account. The agent’s output will be sent as a message directly to you.

    Other delivery options:

    • Slack — posts to a channel or DM
    • Email — sends to any address
    • Webhook — posts a JSON payload to your endpoint
  6. Save and test

    Click Save Job, then click Run Now to trigger the job immediately without waiting for the scheduled time. Check your Telegram — you should receive the briefing within 30–60 seconds.


When your cron job fires, AgenFleet:

  1. Opens a new isolated session for this job (won’t affect your chat history)
  2. Loads the agent’s SOUL file and memory context
  3. Delivers your prompt to the agent
  4. The agent calls web_search, retrieves results, and reasons through them
  5. Produces the formatted output
  6. Sends it to your Telegram
  7. Closes the isolated session

The session is discarded after delivery. Your main chat sessions with the agent are untouched.


After your first run, you can refine:

  • Adjust the prompt — if the output is too long, too short, or missing context you want
  • Change the schedule — edit the cron expression any time
  • Add more delivery targets — send to both Telegram and Slack simultaneously
  • Chain agents — use a webhook delivery to trigger a second agent that acts on the first agent’s output