Agent Events
A simple, open format for giving agents scheduled workflows and event-driven automation.
Agent Events are folders containing event definition that agents can discover and execute automatically.
Skills and tools define what agents can do.
Events define when those capabilities are applied.
Why Agent Events?
Agents are increasingly capable, but typically operate in request-response mode without the ability to schedule future actions or maintain long-running workflows. Agent Events solve this by giving agents a standardized way to define, discover, and execute time-based and event-driven workflows.
For event authors
Build scheduled workflows once and deploy them across multiple agent products.
For compatible agents
Support for events lets end users give agents autonomous, scheduled capabilities out of the box.
For teams and enterprises
Capture organizational workflows in portable, version-controlled, auditable packages.
What can Agent Events enable?
- Scheduled workflows: Daily reports, weekly summaries, monthly reviews - all automated on a cron-like schedule.
- Proactive monitoring: Check system health, scan for issues, send alerts when thresholds are exceeded.
- Recurring tasks: Team status emails, standup reminders, ticket triage - consistent and on time.
- Event-driven automation: React to external triggers like CI/CD failures, customer tickets, or calendar events.
- Interoperability: Reuse the same event across different event-compatible agent products.
Protocol and Format Synergy
Agent Events works seamlessly with the modern AI agent protocol stack:
define when these specialized tasks should be triggered
provide how agents perform specialized tasks
connects agents to external resources
connects agents to other agents (delegation, collaboration)
This creates a complete ecosystem where agents can operate autonomously (Events), leverage expertise (Skills), access external resources (MCP), and coordinate with other agents (A2A).
Event Types & Trigger Mechanisms
Every event answers a fundamental question: "When should it be worth considering an action?" Time-based scheduling is one answer, but Agent Events supports multiple trigger mechanisms to enable sophisticated automation patterns.
1. Time-Based Events
Trigger: Scheduled time intervals
Use cases: Daily reports, health checks, reminders
2. Manual Events
Trigger: Explicit user invocation
Use cases: One-off reports, testing workflows
3. Event-Triggered Events
Trigger: Completion of another event
Use cases: Pipelines, notification chains, workflows
4. State-Change Events
Trigger: Observable state threshold crossing
Use cases: Resource alerts, monitoring, detection
5. External Signal Events
Trigger: Webhooks, API callbacks, MCP signals
Use cases: CI/CD integration, notifications, updates
6. User-Activity Events
Trigger: User presence, behavior, interactions
Use cases: Greetings, reminders, context automation
7. Absence Events
Trigger: Expected signal does not occur
Use cases: Dead-man switches, SLA monitoring
8. Composite Events
Trigger: Multiple conditions satisfied
Use cases: Complex alerts, workflow triggers
Design Principle: Each event type maintains simplicity by answering a single question about timing or conditions. Complexity emerges from composition, not from individual event definitions. This approach preserves agent autonomy while enabling sophisticated automation patterns.
Example Events
Daily Standup Report
Automatically compile team updates from Slack, JIRA, and GitHub and send a formatted summary every morning at 9 AM.
0 9 * * MON-FRI
Code Review Reminder
Check for stale pull requests every 4 hours and send reminders to assigned reviewers with context and priority.
0 */4 * * *
System Health Check
Monitor application metrics, analyze logs, and alert on-call engineers if anomalies are detected.
*/15 * * * *
Weekly Sprint Summary
Aggregate sprint metrics, generate charts, and email stakeholders with progress and blockers every Friday afternoon.
0 16 * * FRI
Specification
This document defines the Agent Events format.
Directory Structure
An event is a directory containing at minimum an EVENT.md file:
event-name/
└── EVENT.md # RequiredYou can optionally include additional directories such as scripts/, references/, skills/ and assets/ to support your events.
EVENT.md Format
The EVENT.md file must contain YAML frontmatter followed by Markdown content.
Frontmatter (required)
---
name: event-name
description: A description of the event, its purpose and when it should be triggered.
---Body Content
The Markdown body after the frontmatter contains the event instructions. There are no format restrictions.
Recommended sections:
- Step-by-step instructions
- Examples of inputs and outputs
- Common edge cases
Optional Directories
scripts/
Contains executable code that agents can run. Scripts should:
- Be self-contained or clearly document dependencies
- Include helpful error messages
- Handle edge cases gracefully
Supported languages depend on the agent implementation. Common options include Python, Bash, and JavaScript.
references/
Contains additional documentation that agents can read when needed:
- REFERENCE.md - Detailed technical reference
- FORMS.md - Form templates or structured data formats
- Domain-specific files (finance.md, legal.md, etc.)
Keep individual reference files focused. Agents load these on demand, so smaller files mean less use of context.
skills/
Contains a specific skill or skills that can be used when the event happens.
assets/
Contains static resources:
- Templates (document templates, configuration templates)
- Images (diagrams, examples)
- Data files (lookup tables, schemas)
Future Vision: Universal EventBus
Agent Events is designed with a forward-looking architecture in mind—a universal EventBus that enables seamless event propagation and subscription across distributed agent systems.
🔄 Event Propagation Layer
Imagine agents publishing events to a shared EventBus infrastructure, enabling real-time event distribution across organizational boundaries. Events become first-class citizens that any authorized agent can subscribe to and react upon.
🌐 Decentralized Event Mesh
A federated network where agents discover and subscribe to event streams from multiple sources—internal systems, third-party services, and other agents—creating an interconnected automation ecosystem.
🔐 Secure Event Distribution
Built-in authentication, authorization, and audit trails ensure events are delivered only to permitted subscribers while maintaining complete transparency of event flows and execution chains.
📡 Event Stream Analytics
Real-time monitoring and analytics of event patterns, enabling intelligent routing, anomaly detection, and optimization of automation workflows at scale.
The Vision: Transform isolated agent workflows into a coordinated, event-driven ecosystem where agents autonomously collaborate, react to real-world signals, and orchestrate complex multi-step processes—all through standardized event interfaces. This is the future of autonomous AI systems.
Open Development
The Agent Events format is open-source and open to contributions from the broader ecosystem. Build on it, extend it, and help shape the future of event-driven AI agents.
Questions or ideas? Contact: krisihadjiev@gmail.com