CodeAtlas MCP Enterprise
Ultra-lightweight, local-first MCP server for AI-powered code intelligence, providing AST-based analysis and 20+ tools while ensuring zero data leakage.
README
<div align="center">
CodeAtlas MCP Enterprise
Enterprise-Grade, Local-First MCP Server for AI-Powered Code Intelligence
CodeAtlas MCP Enterprise is an ultra-lightweight, local-first Model Context Protocol (MCP) server that securely indexes your codebase, performs deep AST-based analysis, and provides 20+ intelligent tools for AI code assistants. Your source code never leaves your machine.
</div>
π Table of Contents
- Why CodeAtlas MCP?
- Features
- Architecture Overview
- Quick Start
- Authentication
- AI Editor Integration
- MCP Tools Reference
- Security Model
- Multi-Tenant Mode
- Environment Configuration
- How It Works
- License
π― Why CodeAtlas MCP?
AI code assistants are powerful β but they work best with context. CodeAtlas gives them X-ray vision into your codebase by:
- π Deep parsing β Understands JavaScript, TypeScript, Python, and PHP at the AST level
- π§ Persistent memory β Retains insights across conversations via Dreaming Memory
- π Zero data leakage β All parsing happens locally, no source code ever transmitted
- β‘ Blazing fast β Full codebase analysis in seconds, incremental re-indexing
- π Universal compatibility β Works with any MCP-compatible editor (Cursor, Claude, VS Code, Windsurf, Copilot)
β¨ Features
| Feature | Description |
|---|---|
| π Local-First Parsing | AST analysis runs entirely on your machine β zero code uploaded |
| π MCP Protocol | Works with all MCP-compatible AI editors |
| π Auto Workspace Discovery | Automatically finds projects in your workspace |
| π Multi-Language AST | JavaScript, TypeScript, Python, PHP with deep dependency resolution |
| π§ Dreaming Memory | Persistent AI memory with vector search for cross-session context |
| π Multi-Tenant Isolation | Isolate projects by workspace with sandbox boundaries |
| π API Key Auth | Secure communication via cryptographic hash verification |
| β‘ Incremental Indexing | Only re-parses changed files for near-instant updates |
| ποΈ Knowledge Graph | Visualize modules, classes, and functions as an interactive graph |
| π Remote Sync | Optionally sync metadata to CodeAtlas Enterprise via HTTPS |
| π Code Metrics | LOC, complexity scores, function counts per project |
| π‘οΈ Security Scanner | Detect hardcoded secrets, unsafe functions, SQL injection |
| π Real-time Watching | Auto re-index on file changes via chokidar |
π Architecture Overview
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β Your Local Machine β
β β
β ββββββββββββ ββββββββββββββ ββββββββββββββββββ β
β β Source βββββΆβ AST βββββΆβ MCP Server β β
β β Code β β Parser β β (this tool) βββββΌβββΆ AI Editor
β β (JS/TS/ β β (local) β βββββββββ¬βββββββββ β
β β PY/PHP)β ββββββββββββββ β β
β ββββββββββββ β β
β βΌ β
β ββββββββββββββββββββ β
β β Dreaming Memory β β
β β (optional: sync) β β
β ββββββββββ¬ββββββββββ β
βββββββββββββββββββββββββββββββββββββββββββββΌββββββββββββββββ
β HTTPS (optional)
βΌ
CodeAtlas Enterprise Server
π Quick Start
Prerequisites
- Node.js v18.0.0 or higher (v20+ recommended)
Install Globally
npm install -g codeatlas-enterprise
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.
π Authentication
Provide your API Key in one of these ways:
-
Environment Variable:
export CODEATLAS_API_KEY="your_api_key_here" -
CLI Argument:
codeatlas-mcp --apiKey="your_api_key_here" -
Local
.envFile (in the directory where you run the command):CODEATLAS_API_KEY=your_api_key_here
π AI Editor Integration
Cursor
Add to ~/.cursor/mcp.json or project-level .cursor/mcp.json:
{
"mcpServers": {
"codeatlas": {
"command": "codeatlas-mcp",
"args": ["--apiKey", "YOUR_API_KEY_HERE"]
}
}
}
Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"codeatlas": {
"command": "codeatlas-mcp",
"args": ["--apiKey", "YOUR_API_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
--apiKeyargument. The local MCP tools (analysis, search, graph) work fully offline.
π MCP Tools Reference
CodeAtlas MCP exposes 20+ tools organized into categories:
Analysis & Indexing
| Tool | Description |
|---|---|
analyze |
Trigger full AST analysis of the current project |
get_project_structure |
Get entities tree (modules, classes, functions, variables) |
get_file_entities |
List all entities defined in a specific file |
Code Exploration
| Tool | Description |
|---|---|
search_entities |
Search for functions, classes, modules by name (fuzzy) |
code_search |
Search source file contents for any text or regex |
get_file_content |
Read file contents with line numbers |
Dependency & Impact Analysis
| Tool | Description |
|---|---|
get_callers |
Find all functions/callers that reference a symbol |
get_callees |
Find everything a function/module imports or calls |
impact_analysis |
Full blast radius: callers + callees + test files |
get_dependencies |
Get import/call/containment/implements relationships |
Visualization & Diagrams
| Tool | Description |
|---|---|
generate_system_flow |
Mermaid flowchart of module architecture |
generate_feature_flow_diagram |
Mermaid sequence/flow diagram for a feature |
trace_feature_flow |
Ordered call chain from entry point to database |
Memory & Persistence
| Tool | Description |
|---|---|
query_dream_memories |
Semantic vector search across past AI memories |
save_dream_memory |
Persist an AI insight or observation for future sessions |
get_system_memory |
Retrieve business rules and change logs |
sync_system_memory |
Save business rules or change descriptions |
Security & Architecture
| Tool | Description |
|---|---|
scan_enterprise_vulnerabilities |
Scan all projects for hardcoded secrets, unsafe functions, SQL injection |
detect_architectural_smells |
Detect circular dependencies, God objects, dead code |
Project Operations
| Tool | Description |
|---|---|
list_projects |
List all discovered and indexed projects |
refresh_projects |
Re-scan directories for new or removed projects |
get_project_insights |
AI-generated refactoring and maintainability suggestions |
π Security Model
π Local-First by Design
CodeAtlas MCP Enterprise follows a zero-trust, local-first architecture:
-
Parsing is local β All source file reading, AST generation, and relationship mapping happens on your machine. No source code is ever uploaded.
-
No credentials embedded β The package contains zero database passwords, Firebase configs, or private server keys. All remote communication uses standard HTTPS with Bearer token auth.
-
Encrypted sync β If you enable remote sync, metadata is transmitted over HTTPS. The server authenticates via cryptographic hash of your API key.
π What Gets Sent (When Sync is Enabled)
Only structural metadata is transmitted:
- File paths and names (relative to project root)
- Function/class/module names and line numbers
- Import/export relationships
- Analysis statistics (file count, LOC, complexity)
Raw source code, credentials, and proprietary logic are never transmitted.
π Multi-Tenant Isolation
When multi-tenant mode is enabled:
- Each tenant's projects are isolated in separate sandbox directories
- Path traversal attacks are blocked by strict boundary validation
- Memory and analysis data are scoped per-tenant
π Multi-Tenant Mode
Enable tenant isolation via environment variables:
CODEATLAS_MULTI_TENANT=true
CODEATLAS_PROJECTS_ROOT=./tenants
Each tenant's projects live in ./tenants/{tenantId}/, with strict path-boundary enforcement.
π Environment Configuration
| Variable | Default | Description |
|---|---|---|
CODEATLAS_API_KEY |
β | API key for authenticating with remote server |
CODEATLAS_API_URL |
https://your-server.com/api |
Remote CodeAtlas server URL |
CODEATLAS_MULTI_TENANT |
false |
Enable multi-tenant isolation |
CODEATLAS_PROJECTS_ROOT |
./tenants |
Root directory for tenant sandboxes |
CODEATLAS_PROJECT_DIR |
process.cwd() |
Default project path |
NODE_ENV |
production |
Environment mode |
βοΈ How It Works
- Start β Run
codeatlas-mcpin your project directory or point it with--projectDir - Auto-Discover β The server scans for projects by detecting
package.json,pyproject.toml,composer.json - AST Parse β Each source file is parsed into an Abstract Syntax Tree
- Build Graph β Modules, classes, functions, and their relationships form a Knowledge Graph
- Serve MCP β AI editors query the graph through 20+ MCP tools
- Dream β Insights persist across sessions via Dreaming Memory (optional remote vector store)
π License
MIT Β© 2026 Giau Phan
π Related Projects
- CodeAtlas AI β Full enterprise server with Oracle 26ai memory, dashboard, security scanner
- npm package β Install via npm
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.