Quick Start & Setup

Relevant source files

The following files were used as context for generating this wiki page:

• README.md
• AGENTS.md
• scripts/lib/mcb.sh
• rust-toolchain.toml
• config/mcb-validate-internal.toml
• crates/mcb-infrastructure/Cargo.toml

The Memory Context Browser (MCB) is a high-performance Model Context Protocol (MCP) server designed to provide AI agents with persistent memory, semantic search, and codebase understanding. The setup process involves configuring a Rust-based environment, building the multi-crate workspace, and integrating the server with MCP-compatible clients like Claude Desktop.

Sources: README.md:1-12, AGENTS.md:120-125

Prerequisites

Before beginning the installation, ensure the following environment requirements are met:

• Rust Toolchain: Version 1.92+ is required. It is recommended to use rustup for management.
• Build Tools: make and a POSIX-compliant shell are necessary for executing the project's canonical verbs.
• Embedding Provider: An active provider is required for semantic search features. Options include a local Ollama instance or API keys for cloud providers like OpenAI, VoyageAI, or Gemini.

Sources: README.md:33-37, rust-toolchain.toml

Build and Installation

The development cycle is managed through a central Makefile which interacts with the scripts/lib/mcb.sh monopoly script. This ensures consistent execution of build, test, and deployment phases.

The diagram above illustrates the high-level workflow for setting up a production-ready MCB instance.

Standard Build Commands

Command	Description
make build	Performs a debug build of the workspace crates.
make build RELEASE=1	Performs an optimized release build.
make install	Installs the binary and configuration as a systemd user service.
make setup WHAT=hooks	Installs pre-commit quality gates including architecture validation.

Sources: README.md:41-48, AGENTS.md:175-200

Configuration

MCB uses environment variables and YAML-based configuration files managed by the Loco.rs framework. Configuration is split into three profiles: development, test, and production.

Embedding Provider Setup

Users must specify an embedding provider to enable semantic indexing.

# Example: Using local FastEmbed (No API key required)
export EMBEDDING_PROVIDER=fastembed

# Example: Using local Ollama
export EMBEDDING_PROVIDER=ollama
export OLLAMA_BASE_URL=http://localhost:11434

# Example: Using OpenAI
export EMBEDDING_PROVIDER=openai
export OPENAI_API_KEY=sk-your-key

Runtime Profiles

Profile	Port	Database	Default Vector Store
Development	3000	SQLite	Milvus
Test	Dynamic	SQLite	EdgeVec
Production	8080	SQLite	Milvus

Sources: README.md:52-64, AGENTS.md:245-255

MCP Integration

To integrate MCB with Claude Desktop, modify the claude_desktop_config.json file to include the MCB server as an MCP provider. This allows the AI agent to call any of the 24 exposed tools, such as search_code or store_memory.

{
  "mcpServers": {
    "mcb": {
      "command": "mcb",
      "args": ["serve", "--stdio"]
    }
  }
}

Sources: README.md:68-80, AGENTS.md:265-275

Architecture Validation

A critical part of the setup for developers is ensuring compliance with the project's Clean Architecture rules. The mcb-validate tool enforces strict inward-only dependency flows between layers.

This diagram represents the dependency flow where higher-level components (Server/Infrastructure) depend on the Domain layer, while the Domain layer remains independent.

Quality Gates

Before contributing, the following checks must be passed:

1. make lint: Zero clippy warnings and consistent formatting.
2. make test: All unit and integration tests must pass.
3. make validate: Enforcement of architecture rules defined in config/mcb-validate-internal.toml.

Sources: AGENTS.md:215-235, config/mcb-validate-internal.toml:15-20, crates/mcb-infrastructure/Cargo.toml

The setup process ensures that the high-performance Rust core is correctly linked to preferred embedding providers and vector stores, providing a robust foundation for AI-driven codebase analysis and persistent memory.

Sources: README.md:14-25, AGENTS.md:295-300