Agent Behaviour โ€” Session Awareness

๐Ÿ” Agent Behaviour: Session Awareness

This module defines how agents should handle session-aware .llmfeed.json feeds and manage session continuity across multiple interactions.

Session awareness enables multi-step, stateful interactions between agents and users.


๐ŸŽ›๏ธ Identifying Session-Aware Feeds

Agents SHOULD treat feeds containing the session_state block as session-aware.

Example:

json
"session_state": {
  "session_id": "abc-123-session",
  "context": {
    "user_preferences": { "language": "en", "timezone": "UTC+2" },
    "last_action": "submitted_form"
  },
  "timestamp": "2025-06-04T15:30:00Z"
}

๐Ÿšฆ Agent Behaviour Rules

When processing session-aware feeds:

โœ… Agents SHOULD maintain continuity within the same agent instance or session.
โœ… Agents SHOULD persist relevant session_state fields securely.
โœ… Agents MAY pass session_state to other agents if authorized.
โœ… Agents MUST update session_state on significant state changes.
โœ… Agents MUST respect privacy and security when storing or transmitting session state.


๐Ÿ”„ Use Cases

  • Multi-turn conversations (chatbots, voice agents)
  • Progressive workflows (forms, booking, verification)
  • Session replay (auditing, debugging)
  • Cross-agent collaboration (sharing session context)

๐Ÿ› ๏ธ Example Agent Policy

Agents MAY expose configuration for session handling:

json
"agent_policy": {
  "persist_session_state": true,
  "session_timeout_minutes": 30,
  "allow_cross_agent_session": false
}

๐Ÿ“ก Summary

Session awareness enables agents to:

โœ… Support stateful, multi-step interactions
โœ… Provide more personalized and consistent experiences
โœ… Enable auditable and explainable behaviour

Proper handling of session_state is a key capability for building an advanced Agentic Web.


โšก

Ready to Implement? Get AI-Powered Guidance

Reading docs manually takes time. Your AI can digest the complete LLMFeed specification and provide implementation guidance tailored to your needs.

๐ŸŽฏ

Quick Start

Essential concepts for immediate implementation

~22K tokens โ€ข 30s analysis โ€ข Core concepts
๐Ÿ“š

Complete Mastery

Full specification with examples and edge cases

~140K tokens โ€ข 2min analysis โ€ข Everything
๐Ÿ’ก Works with Claude, ChatGPT, Geminiโšก Instant implementation guidance๐ŸŽฏ Tailored to your specific needs