Rejoining... Retrying in s Connection lost Session paused Resume failed
Ariadne User Documentation
Documentation for how to use the Ariadne platform

goals-and-schedules

Last updated 13 September 2026

Goals & Schedules

Goals and Schedules let you set agents to work autonomously — either pursuing long-term objectives over time, or running on a fixed timetable.


Goals

What Are Agent Goals?

A goal is a persistent, long-term objective assigned to an agent. Unlike a one-off task in a conversation, a goal is something the agent tracks and works toward continuously.

Examples:

  • "Monitor news about our competitors and update a briefing weekly"
  • "Keep the project documentation up to date as the codebase changes"
  • "Every morning, check my inbox and flag anything that needs a response today"

Creating a Goal

  1. Go to Settings → Goals
  2. Click New Goal
  3. Enter:
    • Title – a short name for the goal
    • Description – a detailed explanation of what the agent should work toward
    • Agent – which agent is responsible for this goal
  4. Click Save

Goals are stored persistently and associated with the assigned agent. The agent can reference its goals when determining what to do during a scheduled or triggered run.

Monitoring Goal Progress

Goals can have progress notes — annotations the agent adds as it makes progress. You can view these from the Goals page to understand what the agent has done and what remains.

Agent-Derived Goals

Agents don't only work on goals you give them. During an overnight review, an agent reflects over your recent conversations and its existing goals and may define one goal of its own per person it works with — for example, noticing you keep asking about a topic and setting itself the goal of staying on top of it.

  • Agent-derived goals appear on the Goals page with an agent-derived badge.
  • They are always lower priority than the goals you set — the agent only works on them once your own goals have had their turn, and only in a small, budgeted amount of idle time each day.
  • You stay in control: edit them, change their success criteria, or delete them just like any other goal. An agent will only ever keep one live self-defined goal per person at a time.

Schedules

What Are Schedules?

Schedules are time-based triggers that automatically start an agent conversation at a specified time or interval. Combined with goals, they allow fully autonomous agent operation.

Examples:

  • Run a briefing agent every weekday at 7:00 AM
  • Check the inbox and triage emails every hour
  • Generate a weekly project summary every Friday at 5:00 PM
  • Run a data monitoring agent every 15 minutes

Creating a Schedule

  1. Go to Settings → Schedules
  2. Click New Schedule
  3. Enter:
    • Name – a descriptive name
    • Agent – which agent to trigger
    • Conversation – the conversation in which the agent will run (or create a dedicated conversation)
    • Schedule – the timing, in cron syntax or a simple interval:
      • Interval: every N minutes/hours/days
      • Cron: e.g. 0 7 * * 1-5 (weekdays at 7 AM)
    • Prompt – the message to send to the agent when triggered (can reference goals)
  4. Click Save

Cron Syntax Reference

Expression Meaning
0 7 * * * Every day at 7:00 AM
0 7 * * 1-5 Weekdays at 7:00 AM
*/15 * * * * Every 15 minutes
0 9 * * 1 Every Monday at 9:00 AM
0 17 * * 5 Every Friday at 5:00 PM

Managing Schedules

  • Enable/Disable – toggle a schedule without deleting it
  • Run Now – trigger the schedule immediately for testing
  • View History – see logs of past scheduled runs and their results
  • Delete – remove a schedule permanently

Combining Goals and Schedules

The most powerful pattern is to combine a goal with a schedule:

  1. Create a goal describing the ongoing objective
  2. Create a schedule that triggers the agent regularly
  3. In the schedule prompt, reference the goal: "Review progress on your active goals and take the next action"

The agent will check its goals, determine what needs to be done, and act accordingly — then update the goal's progress notes.


Use Cases

Scenario Goal Schedule
Morning digest Monitor topics of interest Every weekday at 7 AM
Inbox triage Keep inbox organised Every hour during business hours
Weekly report Track project progress Every Friday afternoon
Competitive monitoring Track competitor news Daily
System health check Monitor service status Every 15 minutes

Success Criteria, Verification & Backoff

Every goal can carry success criteria — a measurable definition of done, edited directly on the goal card:

  • During autonomous work sessions the agent must record concrete progress against the criteria. Sessions that end without recorded progress count as failures.
  • When the agent believes the goal is finished it requests completion, and an independent evaluation checks the claim against the success criteria before the goal actually closes. Rejected claims return feedback and the goal stays open.
  • Goals that repeatedly fail to progress back off (waiting longer between attempts) and are eventually blocked with an explanation — you'll be notified, and editing the goal (or clicking Unblock and retry) puts it back to work. Clearer success criteria usually fix a blocked goal.

While you're actively using an agent, autonomous goal work drops into a low-intensity background lane (small, safe steps only) so it never competes with your conversation; full-depth sessions run when the agent is idle.


Autonomy Activity Feed

Everything the system does on its own initiative — goal sessions, plan steps, scheduled events, watcher observations, notifications, and overnight optimisation passes — is recorded and shown at Tools → Autonomy Activity. Filter by category, and use it to understand exactly what your agents did while you were away.


Watchers

Watchers let agents notice things between conversations. Each watcher polls a source on an interval and turns anything new into an observation agents can read during goal work (and you can review in the activity feed):

  • RSS/Atom feeds — each new item becomes an observation.
  • Web page changes — an observation fires when a watched page's content changes.

Observations are deduplicated, and a newly created watcher starts from a baseline — it won't flood you with everything that already existed.


Proactive Contact Preferences

Agents can proactively reach out — when a goal completes or gets blocked, or when they judge something is worth telling you. You control this from Profile → Proactive Contact:

  • Channels — in-app notifications and/or email
  • Minimum urgency — suppress anything below Low / Normal / High / Critical
  • Quiet hours — a do-not-disturb window in your timezone (critical alerts still get through)
  • Daily limit — a cap on proactive notifications per day

Suppressed notifications aren't lost — they still appear in the Autonomy Activity feed.