Agent Events

Schedule the Future with Event-Driven AI Agents

Agent Events is an open format for giving AI agents scheduled workflows and event-driven automation. 🌐 Website: agentevents.io

Agent Events are folders containing event definitions, scripts(optional),references(optional), assets(optional) and schedules that agents can discover and execute automatically. They enable agents to:

• Run scheduled workflows - Daily reports, weekly summaries, recurring tasks
• Monitor proactively - System health checks, alert thresholds, anomaly detection
• Automate recurring tasks - Team communications, ticket triage, data processing
• React to triggers - CI/CD events, external webhooks, calendar notifications

Most AI agents operate in request-response mode without the ability to schedule future actions or maintain long-running workflows. Agent Events solve this by providing:

• Time-based scheduling - Cron-format schedules for precise timing control
• Skill integration - Compose complex workflows from reusable Agent Skills
• State management - Events can maintain context between executions
• Audit trails - Full execution history and logging
• Portability - Write once, deploy across multiple agent platforms

# events/team-status-mail/event.yaml
name: team-status-mail
description: Send daily status email to the team
schedule: "0 9 * * MON-FRI"  # Every weekday at 9 AM
enabled: true

skills:
  - mail-compose
  - team-context

instruction: |
  Draft and send a brief status email to the team with:
  - Yesterday's key accomplishments
  - Current blockers
  - Today's priorities
  
  Use the team distribution list from team-context skill.

That's it. The agent discovers this event, loads the referenced skills, and executes on schedule.

events/
└── <event-name>/
    ├── EVENT.MD         # Event definition (required)
    ├── scripts/            # Optional: bundled scripts
    ├── references/         # Optional: bundled references
    ├── skills/             # Optional: bundled skills
    │   └── <skill-name>/
    │       └── SKILL.md
    └── assets/             # Optional: templates, data files
        └── templates/

The EVENT.md file must contain YAML frontmatter followed by Markdown content:

---
name: event-name
description: Brief description of what this event does
schedule: "0 9 * * MON-FRI"
enabled: true
---

# Event Instructions

Natural language instructions for the agent describing:
- What task to perform
- When to execute (schedule above)
- What skills or tools to use
- Expected outputs or actions

## Skills

Optional reference to skills (from bundled skills/ directory or global):
- skill-name-1
- skill-name-2

## Additional Context

Any additional context, examples, or guidelines for execution.

Uses standard cron syntax:

┌───────────── minute (0 - 59)
│ ┌───────────── hour (0 - 23)
│ │ ┌───────────── day of month (1 - 31)
│ │ │ ┌───────────── month (1 - 12)
│ │ │ │ ┌───────────── day of week (0 - 6) (Sunday = 0)
│ │ │ │ │
* * * * *

Examples:

• */10 * * * * - Every 10 minutes
• 0 9 * * MON-FRI - Weekdays at 9 AM
• 0 */4 * * * - Every 4 hours
• 0 16 * * FRI - Fridays at 4 PM

For a complete working example of an event-driven agent implementation, see:

Event-Driven Agent Example

This repository demonstrates a full agent implementation with event discovery, scheduling, and execution capabilities.

We welcome contributions!

• Website: agentevents.io
• GitHub: github.com/HadjievK/agentevents
• Discussions: GitHub Discussions
• Issues: GitHub Issues

• Agent Skills - Modular capabilities for agents (skills)
• Model Context Protocol - Standard for tool integration (MCP)
• Claude Code - Anthropic's coding agent
• Cursor - AI-powered code editor

MIT License - see LICENSE for details.

• Inspired by Agent Skills format
• Built with Anthropic Claude API
• Uses Model Context Protocol for tool integration
• Authentication via Microsoft Graph Device Code Flow

Made with ❤️ by the Agent Events community

Schedule the future. Automate the present.