codememory
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@codememoryfind all callers of the login function"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
š§ CodeMemory
The Persistent Intelligence Layer for Codebases
Local-first, privacy-obsessed CLI tool and Model Context Protocol (MCP) server for AI agents.
š Official Website & Docs: https://codemem.vercel.app/
ā” Key Highlights
š 100% Local & Private: Zero telemetry, zero external network calls. All intelligence stored locally in SQLite WAL mode.
š Interactive Web Architecture Explorer: Local visual flowchart & memory dashboard with timeline scrubbing and real-time live updates.
ā” Token Efficiency: Minimizes AI prompt token consumption by providing targeted context slices and caller graphs instead of raw file dumps.
š Multi-Language AST Analysis: Extracts functions, classes, interfaces, structs, and imports across TypeScript/JavaScript, Python, Rust, Go, SQL, and more.
š Real-time Live Watcher: Detects code changes instantly with 100ms debouncing and
.gitignoreintegration.š¤ Native MCP Server: Ready to connect with Antigravity, Claude Desktop, Cursor, Copilot, and custom AI agents.
š Architecture Visualizer: Generates live Mermaid diagrams, Markdown documentation, and JSON maps.
Related MCP server: ContextAtlas
š Quick Start
1. Installation
# Clone the repository
git clone https://github.com/EldrexDelosReyesBula/CodeMemory.git
cd CodeMemory
# Install dependencies
npm install
# Build TypeScript
npm run build2. Basic Commands
# Initialize CodeMemory in your repository
npm run dev -- init
# Scan the current codebase and index all symbols
npm run dev -- scan
# Search for any function, class, or symbol
npm run dev -- query authenticateUser
# Extract token-budgeted AI context
npm run dev -- context AuthService --budget 2000
# Launch the interactive local Web Explorer (auto-opens browser)
npm run dev -- web
# or with explicit launch mode
npm run dev -- web --mode auto
npm run dev -- web --mode manual
# Start in background daemon mode
npm run dev -- web --daemon
# Start the live background file watcher (quiet mode available)
npm run dev -- watch
npm run dev -- watch --quiet
# Export Mermaid architecture diagram
npm run dev -- export --format mermaid
# Check repository health and statistics
npm run dev -- statusš¤ Connecting to AI Agents via MCP
CodeMemory provides a full Model Context Protocol (MCP) stdio server.
Available MCP Tools
Tool | Description |
| Search code symbols, signatures, and functions |
| Token-budgeted context with dependencies & callers |
| Upstream imports and downstream dependents |
| Recent file modifications and Git commit logs |
| Generate real-time Mermaid architecture diagrams |
| Overall codebase stats and language breakdowns |
Antigravity & Claude Desktop Configuration
Add the following to your MCP configuration (mcp_config.json or claude_desktop_config.json):
{
"mcpServers": {
"codememory": {
"command": "npx",
"args": ["-y", "@eldrex/codememory", "mcp"],
"env": {}
}
}
}šļø Architecture
CodeMemory/
āāā src/
ā āāā types/ # Domain data types & contracts
ā āāā db/ # SQLite engine with WAL mode & indexes
ā āāā parser/ # Multi-language AST symbol & dependency extractor
ā āāā watcher/ # Real-time debounced file watcher with .gitignore
ā āāā git/ # Git commit and diff history monitor
ā āāā generator/ # Mermaid diagram & Markdown/JSON exporters
ā āāā context/ # Token-budgeted context assembly & symbol ranker
ā āāā mcp/ # Model Context Protocol stdio server
ā āāā cli.ts # CLI commands router
āāā tests/ # Complete Vitest test suiteš Optional Integrations
CodeMemory works 100% standalone out-of-the-box. When used alongside DevDiff, you can leverage optional integration commands:
# Verify synchronization between Git-tracked files and CodeMemory index
codememory devdiff sync
# Check for untracked or unindexed files in workspace
codememory devdiff compare
# Analyze direct and indirect caller dependencies for specific files
codememory devdiff impact src/cli.ts
# View file change records and AST symbols
codememory devdiff explain src/cli.ts
# Start Unified MCP Server (combines CodeMemory and DevDiff tools)
codememory mcp --unifiedš License
MIT License - Copyright (c) 2026 Eldrex Delos Reyes Bula.
This server cannot be deployed
Maintenance
Related MCP Connectors
Codebase graphs, caller impact analysis, and recorded project context for AI coding agents.
shared AI-context layer for teams ā persistent memory your agents search and update over MCP
Search GitHub, npm, PyPI, StackOverflow, ArXiv from one MCP ā built for coding agents.
Hosted code graph over MCP: exact callers, dependencies, and cross-repo blast radius for AI agents.
Related MCP Servers
- AlicenseBqualityDmaintenanceLocal-first codebase context engine that parses code into a ranked dependency graph and serves it to AI tools via MCP for deep structural understanding.515 npm1MIT
- AlicenseNot gradedqualityDmaintenanceEnables AI coding agents to retrieve and manage code context with hybrid search, project memory, and observability via MCP tools.29MIT
- AlicenseNot gradedqualityDmaintenanceServes structured code context via MCP, enabling AI agents to understand codebases with dependency graphs and significantly reduce token usage.22 npmMIT
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to perform hybrid code search, get explanations, analyze relations and impacts, retrieve context packs, and generate documentation across ~45 languages via 17 MCP tools, all powered by a local vector database and LLM.2Apache 2.0