Conversations: Shared Agent Memory across Concurrent Experiences Fresh

Published: January 21, 2026 | Original Post

---

The Conversations API allows agents to handle multiple concurrent conversations while maintaining unified memory and identity.

Key Features

Multiple Concurrent Threads

Previously agents experienced time as a single thread. The new conversations API allows agents to engage in hundreds of parallel conversations simultaneously while retaining their core identity and shared memory.

Unified Memory System

Each conversation is tied to a specific agent, and experiences within a conversation can form memories which transfer across all conversations. This enables powerful organizational use cases where a single agent serves as a knowledge hub across multiple communication channels.

Practical Applications

The system excels in scenarios like Slack integration, where each thread becomes an independent conversation feeding into one agent's shared memory. The agent can search across all conversations to surface relevant context.

Implementation

Agent Development Environment (ADE): Users can create side conversations alongside the default conversation thread within the same interface.

Letta Code: The latest version defaults to using the Conversations API, supporting multiple concurrent sessions with persistent memory.

Design Considerations

Developers can leverage conversations to serve multiple end-users without creating individual agents per user, while optionally isolating memory by maintaining separate agents if needed.