How to Deploy Codebase Memory MCP: Complete Installation Guide

Deploy Codebase Memory MCP by running the one-line installer that downloads the pre-built static binary, executes the bundled install.sh script to configure supported coding agents, and optionally enables the graph visualization UI with the --ui flag.

Codebase Memory MCP is a self-contained indexing and query engine that creates persistent SQLite knowledge graphs of your codebases. According to the DeusData/codebase-memory-mcp source code, deploying this MCP server involves obtaining the binary, running the automated installer, and optionally building from source when customization is required. Once deployed, the system enables 15 MCP tools that integrate with Claude Code, Codex, VS Code, and other supported agents.

The fastest way to deploy Codebase Memory MCP uses the one-line installer hosted in the repository root. This script automatically detects your operating system, downloads the appropriate pre-built static binary, and executes post-install configuration to set up agent integration.

macOS and Linux Installation

Run the installer directly from the repository:

curl -fsSL https://raw.githubusercontent.com/DeusData/codebase-memory-mcp/main/install.sh | bash

The install.sh script performs several critical operations: it copies the binary to a location on your $PATH, strips macOS quarantine attributes when necessary, and auto-detects supported coding agents to inject the required MCP configuration.

Windows Installation

For Windows environments, use the PowerShell installer located at install.ps1:


# Download the installer

Invoke-WebRequest -Uri https://raw.githubusercontent.com/DeusData/codebase-memory-mcp/main/install.ps1 -OutFile install.ps1

# Review the script (optional)

notepad install.ps1

# Unblock and execute

Unblock-File .\install.ps1
.\install.ps1

Building from Source

When you need a custom build or want to inspect the implementation, clone the repository and invoke the build helper. The core implementation resides in internal/cbm/cbm.c, which handles parsing with vendored Tree-Sitter grammars and RPC handling.

git clone https://github.com/DeusData/codebase-memory-mcp.git
cd codebase-memory-mcp
scripts/build.sh

The build process compiles the C source using Makefile.cbm and produces a static binary at build/c/codebase-memory-mcp. To include the optional graph visualization UI, pass the --with-ui flag:

scripts/build.sh --with-ui

Enabling the Graph Visualization UI

The optional UI provides a 3D web interface for exploring your knowledge graph at http://localhost:9749. When using the one-line installer, append the --ui flag:

curl -fsSL https://raw.githubusercontent.com/DeusData/codebase-memory-mcp/main/install.sh | bash -s -- --ui

If you built from source with UI support, start the server manually:

codebase-memory-mcp --ui=true --port=9749

Post-Deployment Configuration

After installation, the binary automatically configures supported agents to communicate with the MCP server. The system creates a persistent SQLite knowledge graph under ~/.cache/codebase-memory-mcp/ and enables all 15 MCP tools.

Runtime Environment Variables

Configure behavior through environment variables documented in docs/CONFIGURATION.md:

  • CBM_CACHE_DIR: Overrides the default cache location
  • CBM_AUTO_INDEX: Enables automatic indexing for new sessions

Set persistent configuration values using the CLI:

codebase-memory-mcp config set auto_index true

Manual Repository Indexing

To index a project immediately after deployment:

codebase-memory-mcp cli index_repository --repo-path /absolute/path/to/your/project

Query the graph using the search command:

codebase-memory-mcp cli search_graph --project my-project \
  --name-pattern '.*Handler.*' --label Function | jq '.results[].name'

Architecture Overview

Understanding the deployment architecture helps troubleshoot issues. The system consists of four main components as implemented in the DeusData/codebase-memory-mcp repository:

  • Binary (codebase-memory-mcp): The core engine located at build/c/codebase-memory-mcp after compilation, responsible for parsing source files and building the SQLite graph.
  • Coordination Daemon: A per-account background process that serializes indexing jobs and watches for file changes.
  • Agent Integration Layer: Auto-detects supported agents and injects MCP configuration, skills, and hooks.
  • Graph UI: Optional 3D visualization component served on port 9749.

Summary

  • Deploy instantly using the one-line installer from install.sh or install.ps1 for your operating system.
  • Build manually by running scripts/build.sh to produce the static binary at build/c/codebase-memory-mcp.
  • Enable visualization by adding the --ui flag during installation or building with --with-ui.
  • Configure agents automatically through the installer, which detects Claude Code, Codex, VS Code, and other supported clients.
  • Manage settings via the config sub-command and environment variables like CBM_CACHE_DIR.

Frequently Asked Questions

What is the fastest way to deploy Codebase Memory MCP?

The fastest method is the one-line installer: curl -fsSL https://raw.githubusercontent.com/DeusData/codebase-memory-mcp/main/install.sh | bash. This downloads the pre-built static binary, places it on your $PATH, and automatically configures supported coding agents without manual intervention.

Where does Codebase Memory MCP store its knowledge graph data?

The system creates a persistent SQLite knowledge graph under ~/.cache/codebase-memory-mcp/ by default. You can override this location by setting the CBM_CACHE_DIR environment variable or using the config sub-command to modify runtime settings documented in docs/CONFIGURATION.md.

How do I deploy the graph visualization UI alongside the MCP server?

Append the --ui flag when running the installer: curl -fsSL .../install.sh | bash -s -- --ui. Alternatively, when building from source, execute scripts/build.sh --with-ui and then start the server with codebase-memory-mcp --ui=true --port=9749 to access the 3D visualization interface.

Can I deploy Codebase Memory MCP on Windows without WSL?

Yes, the repository includes install.ps1, a native PowerShell installer for Windows. Download the script using Invoke-WebRequest, unblock it with Unblock-File, and execute it to deploy the binary and configure Windows-based coding agents automatically.

Have a question about this repo?

These articles cover the highlights, but your codebase questions are specific. Give your agent direct access to the source. Share this with your agent to get started:

Share the following with your agent to get started:
curl -s "https://instagit.com/install.md"

Works with
Claude Codex Cursor VS Code OpenClaw Any MCP Client

Maintain an open-source project? Get it listed too →