Skip to main content

Jobs

Scheduled Jobs automatically execute agents at specified intervals, enabling automation of repetitive workflows, regular data collection, monitoring tasks, and scheduled reporting within the ToothFairyAI platform.

There are two job types:

  • Task — Executes a specific prompt on a schedule. The agent receives a predefined prompt and performs the exact action specified.
  • Routine — Performs an autonomous heartbeat check on a schedule. The agent probes its tools to identify opportunities, plans actions, and executes them autonomously. Requires Agentic 2.0 (Agentic Tooling) and Long Term Memory to be enabled on the agent.

Jobs can be created and managed from the following menu: Settings > Jobs > Create Job

Create a job​

Basic Information​

  1. Click on the Create Job button.
  2. Job Name - Assign a unique name to identify the job.
  3. Description - Provide a meaningful description of what the job accomplishes.
  4. Job Type - Select the type of scheduled job:

Task​

Execute a specific task on a schedule. The agent receives a predefined prompt and performs the exact action specified. Ideal for deterministic, repeatable workflows such as generating reports, sending notifications, or running data checks.

Routine​

Perform an autonomous heartbeat check on a schedule. The agent follows a three-phase protocol on each execution:

  1. Sensing — The agent probes its mapped tools (knowledge base, internet search, CRM, email, etc.) to identify opportunities, anomalies, or items requiring attention.
  2. Planning — The agent evaluates findings and creates an action plan, prioritizing tasks by urgency and impact.
  3. Acting — The agent self-prompts to execute planned tasks via its tools — sending emails, updating CRM records, generating reports, or any action its tools support.

Routines are ideal for open-ended monitoring, proactive outreach, follow-up management, and any scenario where the agent needs to decide what to do rather than follow a fixed instruction.

Routine Requirements

Routine jobs have the following requirements:

  • The agent must have Agentic 2.0 (Agentic Tooling) enabled. Only agents with Agentic Tooling appear in the agent dropdown when Routine is selected.
  • The agent must have Long Term Memory enabled. If the selected agent does not have Long Term Memory, you will be prompted to enable it directly from the job creation form.
  • The minimum schedule interval for Routine jobs is 30 minutes.

Execution Configuration​

  1. Agent - Select the agent that will execute the job. When creating a Routine job, only agents with Agentic Tooling enabled are shown.
  2. Prompt - Choose the prompt that will be sent to the agent. Available prompts are retrieved from your saved prompts (see Prompting for more details). For both Task and Routine jobs, the prompt content serves as the instruction set:
    • Task: The prompt is the exact instruction the agent will execute.
    • Routine: The prompt becomes the custom instructions appended to the agent's built-in heartbeat protocol. The agent treats these instructions as highest priority, overriding any default behavior.
  3. Max Planned Actions (Routine only) — Set the maximum number of actions the agent can plan and execute per heartbeat cycle (1–20, default 5). This limits the scope of each routine run to prevent runaway execution.

Schedule Configuration​

  1. Active - Toggle to enable or disable the job execution.

  2. Frequency - Select how often the job should run:

    • Minutes - Runs at the specified minute interval (minimum 30 min for Routine)
    • Hourly - Runs every hour at the specified minute
    • Daily - Runs once per day at the specified time
    • Weekly - Runs once per week on the selected day and time
    • Monthly - Runs once per month on the specified day and time
    • Yearly - Runs once per year on the specified date and time

  3. Hour - Select the hour (0-23) when the job should run (for daily, weekly, monthly, and yearly frequencies).

  4. Minute - Select the minute (0-59) when the job should run.

  5. Timezone - Select the timezone for the scheduled execution time.

  6. Start date - Define when the job should begin running.

  7. End date - Define when the job should stop running (optional).

  8. Click on the Save button to create the job.

Managing jobs​

Once created, jobs appear in the jobs list with the following information:

  • Type — Shows whether the job is a Task or Routine
  • Agent - The agent assigned to execute the job
  • Prompt - The prompt being used
  • Status - Shows whether the job is ACTIVE, INACTIVE, or the current phase for Routines (Sensing, Planning, Acting)
  • Schedule - Displays the frequency and timing

Edit a job​

  1. Click on an existing job from the list.
  2. The edit modal displays the job ID and all configuration options.
  3. Modify any settings as needed.
  4. Click Save to update the job.

Delete a job​

  1. Open the job you want to delete.
  2. Click the Delete button at the bottom of the edit modal.
  3. Confirm the deletion when prompted.

Routine phase status​

Routine jobs display real-time phase status during execution:

StatusMeaning
SensingAgent is probing tools to gather information and identify opportunities
PlanningAgent is evaluating findings and creating an action plan
ActingAgent is executing planned actions via its tools

Phase transitions are tracked as best-effort observational metadata. If no phase is detected in the stream, the status remains at the last known phase and the job continues running normally.

Long Term Memory​

Routine jobs rely on the agent's Long Term Memory to recall past conversations and maintain context across heartbeat cycles. This enables the agent to:

  • Remember previous interactions and outcomes
  • Recognize recurring patterns and anomalies
  • Build institutional knowledge over time
  • Make informed decisions based on historical context
Disabling Long Term Memory

If you disable Long Term Memory on an agent that has active Routine jobs, the system will prevent the change and display a warning listing the affected jobs. You must either disable those Routine jobs or convert them to Task type before Long Term Memory can be disabled.

Enabling Long Term Memory from the job form​

When creating a Routine job and selecting an agent that does not have Long Term Memory enabled, you will see a prompt allowing you to enable it directly — without navigating away from the job configuration. Click Enable Long Term Memory and confirm the action.

Job outputs​

The results from executed jobs are automatically saved and made available in the External chats section under the Agents module. This allows you to:

  • Review historical job execution results
  • Monitor job performance and success rates

Use cases​

Task use cases​

  • Daily Reports - Generate business intelligence reports every morning
  • Data Monitoring - Check for data anomalies or threshold breaches
  • Content Generation - Create regular content like newsletters or summaries
  • System Health Checks - Monitor API endpoints or system status
  • Compliance Checks - Run regular compliance or security audits

Routine use cases​

  • Proactive Follow-ups - Agent checks CRM for stale deals and sends follow-up emails
  • Opportunity Detection - Agent monitors knowledge base and internet for relevant market shifts
  • Customer Success - Agent reviews support tickets and proactively reaches out to at-risk accounts
  • Pipeline Management - Agent identifies deals needing attention and updates CRM records
  • Competitive Intelligence - Agent periodically scans for competitor news and summarizes findings

Best practices​

  1. Test First - Run your agent and prompt combination manually before scheduling to ensure expected results.
  2. Clear Naming - Use descriptive job names that indicate purpose and frequency (e.g., "Morning CRM Follow-up Routine" or "Weekly Compliance Report Task").
  3. Timezone Awareness - Always verify the timezone setting to ensure jobs run at the intended time.
  4. End Dates - Consider setting end dates for temporary jobs to prevent unnecessary executions.
  5. Monitor Outputs - Regularly check the Widgets section to verify job outputs and catch any issues early.
  6. Routine Prompt Design - For Routine jobs, write prompts that describe what to look for and what actions to prioritize, not step-by-step instructions. The agent will autonomously determine the best approach.
  7. Max Planned Actions - Start with a lower max planned actions (3–5) and increase only if needed. Higher values lead to longer execution times.
Execution Limits

Jobs are subject to the same execution limits as manual agent runs, including token limits, execution timeouts, and rate limits based on your subscription tier.

Agent Availability

Only agents that have been previously created and saved in your workspace will appear in the agent selection dropdown. For Routine jobs, only agents with Agentic 2.0 (Agentic Tooling) enabled are shown. Ensure you have configured the necessary agents before creating jobs.