AI Second Brain — Persistent memory for any AI client via MCP.

Automatically store and retrieve Dreams, Memories, Genome (DNA), and Immune Genes across Hermes, Claude Code, Gemini CLI, Cursor, and any MCP-compatible AI client.

🚀 Quick Start

# Install
npm install -g codeatlas-enterprise

# Setup (interactive wizard)
export CODEATLAS_API_KEY="ca_..."
codeatlas init

# Or just run the MCP server (auto-detected by clients)
codeatlas

📋 CLI Commands

| Command | Description | |---------|-------------| | codeatlas | Run MCP server (default, no args) | | codeatlas init | Interactive Second Brain setup wizard | | codeatlas setup | Same as init | | codeatlas doctor | Health check & diagnostics | | codeatlas --help | Show usage | | codeatlas --version | Show version |

codeatlas init — Setup Wizard

Steps through:

1. Authenticate with CodeAtlas Cloud
2. Connect your project
3. Enable AI Second Brain
4. Initialize Dreams, DNA, Immune System
5. Verify synchronization
6. Health check — confirm everything works

codeatlas doctor — Health Check

$ codeatlas doctor

CodeAtlas Second Brain — Health Check
==================================================
  ✓ CODEATLAS_API_KEY (ca_7d94a...)
  ✓ Cloud connection (https://atlas.genrostore.com/)
  ✓ MCP config (Hermes)
  ✓ Auto plugin (Hermes)
  ✓ Dream persistence
  ✓ Genome (DNA)
  ✓ Immune System
==================================================
7/7 checks passed. All systems operational.

Run

# Scan current directory and start MCP server
codeatlas-mcp

# With API key for remote sync
codeatlas-mcp --apiKey="your_api_key_here"

# Point to a specific project directory
codeatlas-mcp --projectDir="/path/to/your/project"

That's it! Your AI editor can now connect to the MCP server running on stdio.

🔑 Authentication

Set your API key as an environment variable:

export CODEATLAS_API_KEY="ca_your_key_here"

Or provide it via the setup wizard's interactive prompt.

3. Local .env File (in the directory where you run the command):

   CODEATLAS_API_KEY=your_api_key_here

🧠 AI Second Brain Features

The Second Brain is a persistent knowledge layer shared across all your AI clients:

| Feature | MCP Tool | Description | |---------|----------|-------------| | Dreams | save_dream_memory | Save persistent memories | | | query_dream_memories | Search memories semantically | | Genome (DNA) | search_genome | Search DNA/genes by relevance | | | get_gene | Get gene by ID | | Immune System | scan_immune_genes | Check for known failure patterns | | | save_immune_gene | Record a failure prevention | | Setup | setup_second_brain | Auto-configure any MCP client | | | check_second_brain_status | Show config status for all clients | | System Memory | sync_system_memory | Generate project memory docs | | | get_system_memory | Retrieve system documentation |

🤖 AI Client Integration

Hermes Agent

Config is automatic via codeatlas init. The auto-retrieval plugin is installed at:

~/.hermes/plugins/codeatlas_second_brain/

The plugin hooks into every conversation turn:

• pre_llm_call: auto-retrieve Dreams + DNA + Immune context
• post_llm_call: auto-save new knowledge as Dreams

No user commands needed. The AI automatically recalls past knowledge and saves new learnings.

Claude Code / Claude Desktop

Add to ~/.claude/claude.json:

{
  "mcpServers": {
    "codeatlas": {
      "command": "npx",
      "args": ["-y", "codeatlas-enterprise"],
      "env": { "CODEATLAS_API_KEY": "ca_your_key_here" }
    }
  }
}

Or run codeatlas init which configures this automatically.

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "codeatlas": {
      "command": "npx",
      "args": ["-y", "codeatlas-enterprise"]
    }
  }
}

Gemini CLI

gemini mcp add codeatlas --command "npx" --args "-y" --args "codeatlas-enterprise" --env CODEATLAS_API_KEY="ca_your_key_here"

VS Code / Windsurf / Copilot

For any MCP-compatible editor, use the same JSON structure:

{
  "mcpServers": {
    "codeatlas": {
      "command": "codeatlas-mcp",
      "args": ["--apiKey", "YOUR_API_KEY_HERE"]
    }
  }
}

Note: If you're running without a remote server, omit the --apiKey argument. The local MCP tools (analysis, search, graph) work fully offline.

🔄 Automatic Second Brain Workflow

When the Hermes auto plugin is active, every conversation follows this pipeline:

User: "Continue implementing the AI Gateway"
                        │
                        ▼
  ┌─────────────────────────────────────┐
  │  pre_llm_call (auto)                │
  │  ├─ query_dreams("AI Gateway")      │  → 5 relevant dreams found
  │  ├─ search_genome("AI Gateway")     │  → 3 DNA genes found
  │  └─ scan_immune("AI Gateway")       │  → Prevention context injected
  └──────────────┬──────────────────────┘
                 │
                 ▼
        LLM Reasoning
   (with Second Brain context)
                 │
                 ▼
  ┌─────────────────────────────────────┐
  │  post_llm_call (auto)                │
  │  └─ save_dream("new knowledge")      │  → Dream auto-persisted
  └──────────────────────────────────────┘

No user said "search", "load", "save", or "memories" — all automatic.

📦 All MCP Tools (30 total)

| Tool | Purpose | |------|---------| | save_dream_memory | ✅ Save a dream memory | | query_dream_memories | ✅ Query dreams by semantic search | | search_genome | ✅ Search CodeAtlas Genome | | get_gene | ✅ Get a specific gene by ID | | scan_immune_genes | ✅ Scan immune system for prevention | | save_immune_gene | ✅ Record a failure pattern | | setup_second_brain | ✅ Auto-configure MCP clients | | check_second_brain_status | ✅ Show Second Brain config | | analyze | ✅ Deep code analysis | | list_projects | ✅ List analyzed projects | | get_file_entities | ✅ Get entities in a file | | get_project_structure | ✅ Get project structure | | code_search | ✅ Search file contents | | get_callers | ✅ Find reverse dependencies | | get_callees | ✅ Find forward dependencies | | impact_analysis | ✅ BLAST radius analysis | | get_dependencies | ✅ Get import/call relationships | | generate_system_flow | ✅ Generate Mermaid flow diagram | | trace_feature_flow | ✅ Trace execution flow | | get_system_memory | ✅ Get system documentation | | sync_system_memory | ✅ Generate memory docs | | get_insights | ✅ AI code insights | | detect_architectural_smells | ✅ Detect circular deps, God objects | | scan_enterprise_vulnerabilities | ✅ Security scan | | project_context | ✅ Project overview | | run_script | ✅ Run npm scripts | | git_changes | ✅ Recent git changes | | generate_feature_flow_diagram | ✅ Feature flow diagrams | | list_resources | ✅ List MCP resources | | read_resource | ✅ Read MCP resource |

🏗 Architecture

┌─────────────────────┐     MCP stdio      ┌──────────────────────┐
│  Hermes Agent       │◄──────────────────►│  codeatlas-enterprise │
│  Claude Code        │                    │  MCP Server           │
│  Gemini CLI         │                    │                       │
│  Cursor             │                    │  CLI:                 │
│  Any MCP client     │                    │  ├─ init, setup       │
└─────────────────────┘                    │  ├─ doctor            │
                                           │  └─ (no args → MCP)  │
                                           │                       │
                                           │  HTTPS (REST)         │
                                           │         │             │
                                           │         ▼             │
                                           │  CodeAtlas Cloud API  │
                                           │  atlas.genrostore.com │
                                           └──────────────────────┘

🔒 Security

• Local-First: Source code is parsed locally. No raw files uploaded.
• API Key Auth: All cloud requests authenticated via CODEATLAS_API_KEY.
• HTTPS Only: All communication encrypted.
• PID Guard: Prevents duplicate MCP server instances.

📄 License

MIT © 2026 Giau Phan

🔗 Related Projects

• CodeAtlas AI — Full enterprise server with Oracle 26ai memory, dashboard, security scanner
• npm package — Install via npm