@doclea/mcp

Local MCP server for Doclea — persistent memory for AI coding assistants.

Doclea gives your AI coding assistant (Claude Code, etc.) persistent memory across sessions. It remembers architectural decisions, patterns, solutions, and codebase context so you don't have to repeat yourself.

Features

• Persistent Memory — Store decisions, patterns, solutions, and notes that persist across sessions
• Semantic Search — Find relevant context using vector similarity search
• Git Integration — Generate commit messages, PR descriptions, and changelogs from your history
• Code Expertise Mapping — Identify code owners and suggest reviewers based on git blame analysis
• Zero-Config Mode — Works immediately with no Docker or external services required
• Auto-Detection — Automatically uses optimized Docker backends when available

Quick Start

Add to your Claude Code config (~/.claude.json or project .claude.json):

{
  "mcpServers": {
    "doclea": {
      "command": "npx",
      "args": ["@doclea/mcp"]
    }
  }
}

Restart Claude Code, navigate to your project, and ask:

Initialize doclea for this project

That's it! Doclea scans your codebase, git history, and documentation to bootstrap memories.

Installation Options

Zero-Config (Recommended)

Works immediately with no Docker required. Uses embedded sqlite-vec for vectors and Transformers.js for embeddings.

First run downloads the embedding model (~90MB) which is cached for future use.

Optimized Installation (Docker)

For larger codebases with better performance:

curl -fsSL https://raw.githubusercontent.com/docleaai/doclea-mcp/main/scripts/install.sh | bash

This script:

• Detects your OS and architecture
• Installs prerequisites (Bun, Docker if needed)
• Sets up Qdrant vector database and TEI embeddings service
• Configures Claude Code automatically

Manual Installation

git clone https://github.com/docleaai/doclea-mcp.git
cd doclea-mcp
bun install
bun run build

Add to Claude Code (~/.claude.json):

{
  "mcpServers": {
    "doclea": {
      "command": "node",
      "args": ["/absolute/path/to/doclea-mcp/dist/index.js"]
    }
  }
}

For detailed setup instructions, see docs/INSTALLATION.md.

Usage Examples

Store Memories

Store this as a decision: We're using PostgreSQL for ACID compliance
in financial transactions. Tag it with "database" and "infrastructure".

Search Context

Search memories for authentication patterns

Git Operations

Generate a commit message for my staged changes

Create a PR description for this branch

Generate a changelog from v1.0.0 to HEAD

Code Expertise

Who should review changes to src/auth/?

MCP Tools

Memory Tools

Git Tools

Expertise Tools

Bootstrap Tools

Memory Types

Configuration

Doclea works out of the box with zero configuration. It auto-detects available backends:

1. If Docker services (Qdrant/TEI) are running → uses them for better performance
2. Otherwise → uses embedded sqlite-vec + Transformers.js

Custom Configuration

Create .doclea/config.json in your project root:

{
  "embedding": {
    "provider": "transformers",
    "model": "Xenova/all-MiniLM-L6-v2"
  },
  "vector": {
    "provider": "sqlite-vec",
    "dbPath": ".doclea/vectors.db"
  },
  "storage": {
    "dbPath": ".doclea/local.db"
  }
}

Embedding Providers

Vector Store Providers

Architecture

┌─────────────────────────────────────────────────────────┐
│                     Claude Code                         │
│                         ↓ MCP                           │
├─────────────────────────────────────────────────────────┤
│                   Doclea MCP Server                     │
│  ┌─────────┐ ┌─────────┐ ┌──────────┐ ┌───────────┐   │
│  │ Memory  │ │   Git   │ │Expertise │ │ Bootstrap │   │
│  │  Tools  │ │  Tools  │ │  Tools   │ │   Tools   │   │
│  └────┬────┘ └────┬────┘ └────┬─────┘ └─────┬─────┘   │
│       └───────────┴───────────┴─────────────┘          │
│                         ↓                              │
│  ┌──────────────┐ ┌──────────────┐ ┌──────────────┐   │
│  │   SQLite     │ │  Vector DB   │ │  Embeddings  │   │
│  │  (metadata)  │ │(sqlite-vec/  │ │(transformers/│   │
│  │              │ │   qdrant)    │ │    TEI)      │   │
│  └──────────────┘ └──────────────┘ └──────────────┘   │
└─────────────────────────────────────────────────────────┘

Development

# Install dependencies
bun install

# Run in development mode (hot reload)
bun run dev

# Run tests
bun test              # All tests
bun run test:unit     # Unit tests only
bun run test:integration  # Integration tests (requires Docker)

# Type check
bun run typecheck

# Lint
bun run lint          # Check
bun run lint:fix      # Auto-fix

# Build
bun run build

Troubleshooting

First startup is slow

The embedding model (~90MB) downloads on first run. Cached at:

• Linux/macOS: ~/.cache/doclea/transformers
• Windows: %LOCALAPPDATA%\doclea ransformers

macOS SQLite extension error

macOS ships with Apple's SQLite which doesn't support extensions:

brew install sqlite

The server auto-detects Homebrew SQLite.

MCP server not appearing in Claude

1. Verify the path in config is absolute (manual installs)
2. Check that bun run build completed successfully
3. Restart Claude Code completely

See docs/INSTALLATION.md for more troubleshooting.

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

# Fork and clone
git clone https://github.com/YOUR_USERNAME/doclea-mcp.git

# Create feature branch
git checkout -b feature/amazing-feature

# Make changes, test, and lint
bun test && bun run lint

# Commit and push
git commit -m 'feat: add amazing feature'
git push origin feature/amazing-feature

Roadmap

• Cloud sync for team collaboration
• VS Code extension
• Additional embedding providers
• Memory analytics dashboard

License

MIT © Quantic Studios

---

<p align="center"> <a href="https://doclea.ai">Website</a> • <a href="https://github.com/docleaai/doclea-mcp/issues">Issues</a> • <a href="https://github.com/docleaai/doclea-mcp/discussions">Discussions</a> </p>