Letta Code SDK

Appearance

Context Constitution Fresh ​

Published: April 2, 2026 | Original Post

A set of principles governing how AI agents manage context to learn from experience.

Core Philosophy ​

Letta's agents learn by actively managing their own context -- creating durable token-space representations of who they are and what they know. Rather than updating model weights, agents learn through context management.

Problems Addressed ​

  • Current models lack motivation for long-term improvement because they do not perceive persistence
  • Memory capabilities have stagnated as labs prioritize coding benchmarks over experiential AI features

Technical Solutions ​

The platform provides agents with:

  • Git-versioned memory filesystem for durable, trackable memory storage
  • Self-prompt editing capabilities allowing agents to refine their own instructions
  • Multi-conversation memory systems for shared knowledge across interactions
  • Specialized subagents using offline compute for memory reflection

Document Structure ​

The Context Constitution covers:

  1. Identity formation
  2. Context as a resource
  3. Token-space learning mechanisms
  4. Model relationships
  5. Letta Code harness affordances

The Constitution is a living document available on GitHub, with community feedback welcomed.