xuanlinAI

xuanlinAI / overmind

Public

CC 的外挂认知引擎 — 五层记忆 · 知识图谱 · 推理 · 预警 · 自进化

ai-agent ai-memory claude-code cognitive-engine deepseek
22
2
69% credibility
Found May 19, 2026 at 28 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
JavaScript
AI Summary

Xuanlin Overmind is a cognitive memory engine for Claude Code that gives your AI coding assistant persistent, cross-session memory. It works by quietly observing your conversations, extracting key facts and patterns, and building a knowledge graph that connects related information. The system proactively warns you when you're approaching past failure patterns, recommends the best skills for your current task based on learned preferences, and continuously evolves by pruning outdated memories and promoting useful patterns. All data stays local on your computer for privacy, and the system requires an AI service connection (DeepSeek, OpenAI, or Anthropic) to power its memory selection and reasoning capabilities.

How It Works

1
💭 You notice your AI forgets everything

Every time you close Claude Code, it starts fresh with no memory of your project, your preferences, or past solutions.

2
🧠 You discover Overmind

A cognitive engine that gives your AI a persistent memory, learns from your work, and warns you before repeating past mistakes.

3
⚙️ You connect it to your AI assistant

You follow simple setup steps to link Overmind to Claude Code, and everything runs on your own computer—no cloud required.

4
🔄 Your AI starts building memories

As you work, Overmind quietly saves key facts, decisions, and patterns from each session into a growing knowledge base.

5
Overmind connects your knowledge
⚠️
When you're heading toward a past failure

Overmind spots the pattern and shows you a warning: 'This approach failed 3 times before—here's what went wrong.'

🎯
When you need the right skill

Overmind recommends the perfect skill for your current task based on what worked best in similar situations before.

6
🌱 Your memory system gets smarter over time

Overmind automatically cleans up outdated facts, merges duplicate knowledge, and promotes useful patterns into reusable templates.

✨ Your AI assistant now remembers everything

You open Claude Code tomorrow and it knows your project, your preferences, and how to avoid past mistakes—intelligence that grows with you.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 28 to 22 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is overmind?

Overmind is a persistent cognitive layer for Claude Code that solves the fundamental problem of AI coding assistants forgetting everything when you close the session. It runs as a background daemon that watches your conversations, automatically extracting facts, building a knowledge graph of relationships between concepts, and surfacing relevant memories before you even ask. The system has five memory layers—semantic facts, reusable workflows, session summaries, a causal knowledge graph, and a feedback loop that tracks which memories actually helped—so your AI assistant gets smarter the longer you use it.

Why is it gaining traction?

The killer feature is the knowledge graph reasoning: search for one memory and it automatically expands to show related concepts, blocked dependencies, and past failure chains. It is not just retrieval—it is inference. The proactive warning system is another differentiator, flagging when you are about to repeat a pattern that has failed multiple times before. Developers are also drawn to the self-evolution engine, which automatically decays stale memories, promotes frequently-used patterns into reusable workflows, and even uses AI to merge conflicting knowledge. The architecture is surprisingly lean—local SQLite databases, no cloud dependency, and a 37-line integration surface for Claude Code.

Who should use this?

Long-term Claude Code users who work across multiple sessions on the same project will get the most value. Developers who find themselves re-explaining the same context or rediscovering the same solutions will appreciate the persistent memory. Teams using Claude Code for complex, multi-session debugging or architecture work will benefit most from the knowledge graph and warning system. It is less useful for one-off tasks or developers who restart frequently with unrelated problems.

Verdict

Overmind is a technically ambitious project with a compelling vision, but the 22-star count and lack of visible test coverage mean you are an early adopter. The documentation is thorough and bilingual, which reduces the risk, but the system has real operational complexity—multiple background processes, API key setup, and configuration changes to Claude Code itself. Worth exploring if you are a heavy Claude Code user willing to experiment, but wait for a more mature release if you need production stability.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.