Appearance
Letta's Next Phase Fresh
Published: March 16, 2026 | Original Post
Letta's pivot toward agent infrastructure focused on memory-first systems.
Core Thesis
Every closed foundation lab today is moving towards developing not just models, but also agents.
Letta Code emerges as the flagship offering -- a model-agnostic agent harness with persistent memory providing frontier agent capabilities including computer use, skills, subagents, and transparent memory systems.
Strategic Improvements
- Memory systems transitioning from specialized database tools to general-purpose computer access via git-backed files (MemFS)
- Sleep-time compute moving to client-side, enabling agents to reflect and improve autonomously
- Agent orchestration expanding beyond messaging to dynamic subagents and skills
- SDK development making systems deployable at scale with memory-first design
Deprecation Timeline
Legacy features being phased out:
| Feature | Status |
|---|---|
| Letta Filesystem | Replaced by actual filesystem access |
Legacy memory tools (core_memory_replace) | Deprecated |
| Templates, hardcoded tool rules | Sunset by mid-April 2026 |
| Server-side MCP integrations | Being phased out |
Mission Statement
Letta builds agents that learn. Agents with persistent memory, real computer access, and infrastructure to improve from their own lived experience.