Claude Subconscious adds memory to Claude Code CLI
Letta AI (formerly MemGPT) released Claude Subconscious, an experimental background agent that provides Anthropic’s Claude Code CLI with long-term memory and autonomous research capabilities across developer sessions.
Persistent state in AI coding agents is the next major hurdle, and Letta is tackling it by decoupling memory from the main chat loop.
- –Background agent observes CLI sessions asynchronously, avoiding workflow latency.
- –Uses the MemGPT architecture to "whisper" relevant context and architectural patterns back to Claude Code.
- –A single "subconscious" agent can bridge multiple projects, sharing developer preferences and global styles.
- –Represents a shift toward multi-agent developer environments where background processes handle routine research and memory management.
DISCOVERED
63d ago
2026-03-25
PUBLISHED
63d ago
2026-03-25
RELEVANCE