Transform your Claude Code into a self-improving AI assistant that remembers context, learns from every interaction, and evolves with your workflow.
Everything you need to turn Claude into a learning, evolving AI companion that grows smarter with every session.
Claude automatically loads domain knowledge at session start and learns from every interaction. No more repeating context.
After significant work, Claude prompts you to capture learnings. Patterns, lessons, and preferences are stored for future sessions.
Automatically analyzes your workflows to detect repeating patterns. Get recommendations for skills to create.
Facts, events, and session logs persist across conversations. Claude remembers what matters to you.
Route tasks to Claude, Codex, or Gemini based on their strengths. Get the best AI for each job.
Run /evolve to analyze patterns, generate skills, and improve the system. Claude gets better over time.
Three simple steps to transform your Claude Code experience.
Get the plugin and run the setup script to configure hooks and create the expertise structure.
Use Claude Code as you normally would. The system automatically tracks patterns, loads expertise, and prompts for learnings.
Run the evolve command periodically to analyze patterns and improve the system. New skills are generated automatically.
Set up Neural Claude Code in your projects in minutes.
Clone the plugin to a permanent location on your machine.
$ git clone https://github.com/brolag/neural-claude-code-plugin
$ cd neural-claude-code-pluginThis configures global hooks and sets up the plugin root.
$ chmod +x scripts/setup-hooks.sh
$ ./scripts/setup-hooks.shAdd this to your shell profile (~/.zshrc or ~/.bashrc).
# Neural Claude Code Plugin
export CLAUDE_PLUGIN_ROOT="$HOME/Sites/neural-claude-code-plugin"Initialize the Neural Claude Code structure in your project.
$ cd your-project
$ mkdir -p .claude/{expertise,memory,scripts,data}
$ mkdir -p .claude/memory/{events,facts,session_logs}Add domain expertise for your project. Start with a base template.
# Project Expertise
domain: your_project_domain
version: 1
last_updated: 2025-01-01
understanding:
project_type: "web-app" # or api, cli, library, etc.
tech_stack:
- typescript
- react
- nodejs
patterns:
- "Components go in src/components/"
- "API routes in src/api/"
lessons_learned: []
open_questions: []
user_preferences: []Define load order and dependencies between expertise files.
version: 1
last_updated: 2025-01-01
load_order:
- project
- domain-specific
confidence:
promotion_threshold: 0.7
deprecation_threshold: 0.3
auto_update:
enabled: trueOverride or extend global hooks with project-specific behavior.
#!/bin/bash
# Project-specific session start hook
# Source global hook first
if [ -f "$CLAUDE_PLUGIN_ROOT/scripts/hooks/session-start.sh" ]; then
source "$CLAUDE_PLUGIN_ROOT/scripts/hooks/session-start.sh"
fi
# Add project-specific initialization here
echo '{"systemMessage": "Project: Your Project Name"}'~/Sites/neural-claude-code-plugin/templates/ for a quick start.
your-project/
├── .claude/
│ ├── expertise/ # Domain knowledge (YAML)
│ │ ├── manifest.yaml # Load order & dependencies
│ │ ├── project.yaml # Project-specific expertise
│ │ └── shared.yaml # Cross-domain knowledge
│ │
│ ├── memory/ # Persistent memory
│ │ ├── events/ # Daily event logs (JSONL)
│ │ │ └── 2025-01-15.jsonl
│ │ ├── facts/ # Stored facts (JSON)
│ │ ├── session_logs/ # Session transcripts
│ │ ├── active_context.md # Current context
│ │ ├── context-cache.json # Hot memory cache
│ │ └── pattern-index.json # Detected patterns
│ │
│ ├── scripts/ # Project hooks
│ │ ├── session-start.sh # On session start
│ │ ├── session-stop.sh # On session end
│ │ └── event-indexer.js # Pattern detection
│ │
│ ├── agents/ # Custom agents
│ ├── skills/ # Reusable skills
│ ├── commands/ # Slash commands
│ ├── data/ # Session state
│ │ └── current-session.json
│ │
│ ├── settings.json # Project settings
│ └── CLAUDE.md # Project instructions
│
└── ... your project filesYAML files storing learned patterns, lessons, and user preferences. Auto-loaded at session start.
Event logs, facts, and session history. Analyzed by /evolve for pattern detection.
Hooks that run on session events. Extend global hooks with project-specific logic.
Custom agents and reusable skills specific to your project.
/evolve
Analyze patterns and improve the system
/remember <fact>
Save a fact to persistent memory
/recall <query>
Search and retrieve from memory
/health
Check system health and maintenance
Real improvements from closing the learning loop.
Join developers who are building with AI that actually learns and improves over time.
Get Started on GitHub