Versions:

  • 0.32.1
  • 0.31.12
  • 0.31.11
  • 0.31.8
  • 0.31.6
  • 0.31.5
  • 0.30.25
  • 0.29.12
  • 0.28.17
  • 0.28.12
  • 0.28.11
  • 0.28.4
  • 0.28.0
  • 0.27.29
  • 0.27.11
  • 0.27.9
  • 0.27.8
  • 0.27.7
  • 0.27.5
  • 0.27.3
  • 0.26.5
  • 0.25.10
  • 0.25.8
  • 0.24.13

Letta Code is a memory-first coding harness developed and published by Letta, currently available at version 0.32.1 as part of a release history spanning 24 versions. Positioned in the AI-assisted software development category, it functions as a coding agent designed specifically for long-lived agents that can learn from experience over time. The core purpose of Letta Code is to move away from the conventional model of independent, disposable coding sessions and instead provide developers with a persisted agent whose accumulated memory carries forward from one interaction to the next. This approach allows the agent to build on prior work, retain context about a project, and improve its usefulness as experience accumulates, rather than starting from scratch each time a new session begins. A defining characteristic of the software is the portability of this agent memory across multiple large language models, including Claude, GPT, Gemini, GLM, Kimi, and others, meaning that users are not locked into a single model provider and can switch or combine models without losing the agent's learned context. This makes Letta Code suitable for use cases such as ongoing software engineering tasks, iterative development on long-running codebases, and workflows where continuity of context is valuable, including scenarios where a development team or individual developer wants an agent that remembers prior decisions, code structure, and established conventions. By treating memory as the foundation of the coding experience, Letta Code addresses a common limitation of session-based coding assistants, where context is lost between interactions and must be manually reconstructed. The active version history of 24 releases, culminating in version 0.32.1, reflects continued development of the harness, and the software's design makes it relevant for developers seeking a persistent, model-flexible coding companion that evolves alongside their projects.

Tags: