Doclea

You run a local MCP server that lets your AI coding assistant remember decisions, patterns, and context across sessions. It provides persistent memory, semantic search, and Git integration so you can automate memories, find related context, and generate meaningful actions from your codebase.

How to use

Set up your client to connect to the local Doclea MCP server and start using memories, searches, and code-aware actions. Start by ensuring the MCP server is available locally. Then configure your MCP client (such as Claude Code) to point at the server, add memories, and invoke actions like memory storage, semantic search, and Git-assisted descriptions.

How to install

Prerequisites you need on your machine before installing:

• Node.js is installed on your system
• bun is installed and available in your PATH
• git is installed
• curl is installed for optional install scripts

Zero-Config (recommended) installs and runs the MCP server with embedded backends. It works immediately with no Docker required.

Install using zero-config: create or update your Claude Code config to include the MCP server reference.

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

Restart Claude Code and in your project, run an action like Initialize doclea for this project to bootstrap memories.

If you prefer an optimized or manual setup, choose one of the following options.

# Optimized installation (Docker-based backends)
curl -fsSL https://raw.githubusercontent.com/docleaai/doclea-mcp/main/scripts/install.sh | bash

This script detects your OS, installs prerequisites as needed, and configures Docker services for vector search and embeddings, then wires Claude Code automatically.

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

Then add the MCP server in Claude Code config as shown in the zero-config example above.

For reference, you can also run a manual setup flow to customize further. It starts from cloning, installing, building, and then pointing your client to the built runtime.

Configuration and usage notes

Key features of the Doclea MCP server include persistent memory storage for decisions, patterns, and notes; semantic search over memories; Git integration for commit messages, PR descriptions, and changelogs; code ownership mapping; and an auto-detect mechanism that uses embedded backends when Docker services are not available.

Memory types you store include decision, solution, pattern, architecture, and note. You can perform semantic searches to locate relevant context and generate actionable items from your history.

Memory and tool overview

The MCP provides a set of tools to manage memories, access and update them, and perform code expertise tasks. Examples include storing a memory, performing a semantic search, getting or deleting memories, and generating Git-related outputs.

Available tools

doclea_store

Store a memory (decision, solution, pattern, architecture, note) across sessions for quick recall and context retention.

doclea_search

Perform semantic search across memories to retrieve relevant context and notes.

doclea_get

Retrieve a memory by its ID for inspection or reuse.

doclea_update

Update an existing memory with new information or corrections.

doclea_delete

Delete a memory from the store.

doclea_commit_message

Generate a conventional commit message from staged changes.

doclea_pr_description

Generate a PR description with context from memory and code history.

doclea_changelog

Generate a changelog between two references.

doclea_expertise

Map codebase expertise and identify bus-factor risks.

doclea_suggest_reviewers

Suggest PR reviewers based on file ownership.

doclea_init

Initialize a project by scanning git history, docs, and code to bootstrap memories.

doclea_import

Import memories from Markdown files or ADRs into memory store.