Zikkaron

Overview

Zikkaron is a sophisticated MCP server that implements a biologically-inspired memory system for Claude Code. It uses computational neuroscience principles including Hopfield networks, predictive coding, and causal discovery to create a persistent memory that survives sessions. The system runs entirely locally on SQLite, avoiding API calls while maintaining exceptional recall performance (86.8% Recall@10). It includes 24 MCP tools for memory management, context reconstruction, and decision tracking across sessions. The system includes advanced features like Hippocampal Replay to preserve context during compaction, automatic action capture, and project seeding for cold starts.

Try asking AI

After installing, here are 7 things you can ask your AI assistant:

• recall

  Retrieves memories associated with the current project name.

  Try:Recall my previous work on the authentication systemTry:What do we know about the project structure?
• get_project_context

  Gets context for the current directory before starting a task.

  Try:Get context for the current directoryTry:What's the current project setup?
• remember

  Stores decisions and outcomes after completing significant work.

  Try:Remember that we chose Redis over MemcachedTry:Store that we fixed the authentication bug
• sync_instructions

  Automatically syncs Claude instructions with the latest CLAUDE.md.

  Try:Sync the latest instructionsTry:Update my memory with current preferences
• install_hooks

  Installs hippocampal replay hooks for a project directory.

  Try:Install hooks for this projectTry:Set up automatic memory consolidation
• anchor

  Anchors important facts to protect them from context compaction.

  Try:Anchor that we're using event sourcingTry:Remember this is a critical constraint
• seed_project

  Seeds a project with foundational memories from structure, configs, and documentation.

  Try:Seed this project with current structureTry:Create initial memories for the codebase

Comparable tools

Installation

pip install zikkaron

{
  "mcpServers": {
    "zikkaron": {
      "command": "zikkaron"
    }
  }
}

## Memory
- On every new session, call `recall` with the current project name
- Before starting any task, call `get_project_context` for the current directory
- After completing significant work, call `remember` to store decisions and outcomes

FAQ

How does Zikkaron handle context compaction in Claude?

Zikkaron implements 'Hippocampal Replay' which drains context before compaction and reconstructs it intelligently after, preserving important decisions and context.

Does Zikkaron require an API key or external services?

No, Zikkaron runs entirely locally on SQLite with a 22MB embedding model, requiring no API calls or external services.

How does Zikkaron perform compared to other memory systems?

Zikkaron significantly outperforms existing systems on benchmarks like BEAM, scoring 0.404 overall (52% improvement over best published system) and excelling in contradiction resolution (+226%), temporal reasoning (+133%), and knowledge updates (+73%).

Compare Zikkaron with