Letta Code Brings Memory to Coding Agents

Letta Code bets that long-lived, stateful agents will outlast session-bound copilots—but self-editing memory creates a new review surface developers cannot ignore.

• –MemFS tracks agent context through Git, making memory inspectable, versioned, and portable. [Context Repositories](https://www.letta.com/blog/context-repositories/)
• –Agents can rewrite memory blocks, skills, prompts, and even parts of their harness through mods, enabling genuine adaptation but increasing operational risk.
• –Cross-device and cross-channel continuity is a meaningful differentiator for teams that want one persistent engineering agent.
• –Model flexibility keeps Letta Code from being locked to a single provider, while its CLI-first workflow fits existing developer environments.
• –The key adoption question is whether teams will review memory changes with the same discipline they apply to code changes. [Independent review](https://mrkeyoor.com/repos/letta-code/)