Claude Soul: a learning system for Claude Code
Developers have introduced Claude Soul, an MCP server that creates long-term memory across sessions. Unlike simple logs, the tool extracts signals from edits and errors, forming behavioral frameworks. The system adjusts its trust in these rules based on experience.
The project author noted unusual effects after 200 sessions. The model built its own memory system, began to argue against bad ideas, and developed a technique for multi-sided analysis without direct prompts.
The tool runs locally and is distributed under the MIT license. This is a significant addition to the earlier /goal mode, as the autonomous agent can now learn from its mistakes across different projects.
Up next
Anthropic doubles Claude Design limits
Anthropic has doubled the token limits for Claude Design on all paid plans. Designers can now generate twice as many prototypes and layouts per week. This is a clear signal: the company is rapidly scaling up its computational resources and is ready to give more power to professional users.
Related stories
- Synapse indexes your codebase locally and feeds it to Claude over MCP
- Introducing Claude Subagent Install Manager
- Claude Code 2.1.319 drops auto mode fallback behind gateways
- Intent-lens spells out Claude Code calls before they run
- Codedict cuts Claude Code tokens 63%, but cost only 26%
- claude-files-pane shows Claude Code’s file reads live
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.