@mempalace/mcp-dev

@mempalace/mcp-dev

Enables AI coding assistants to access grounded, branch-scoped codebase context via semantic search, git tracking, change ledger, and structured feature management with Project Tracks.

Category
访问服务器

README

@mempalace/mcp-dev

An MCP (Model Context Protocol) server that gives AI coding assistants grounded, branch-scoped context about your codebase. It provides semantic code search via my MemPalace fork, git change tracking, a persistent change ledger, TypeScript/Jest dev lifecycle tools, and a built-in Project Tracks (Conductor) methodology for structured feature work.

Features

Feature Description
Semantic code search Natural-language queries over your indexed codebase via MemPalace embeddings
Branch context Current branch, SHA, recent commits — always oriented
Change ledger Append-only JSONL log of every AI-assisted change with full provenance
Dev lifecycle Run tsc typecheck and Jest tests from within the MCP client
Git diff Full or file-scoped diff output
Project Tracks Create and manage markdown-based feature tracks (plan/spec/index/metadata)

Prerequisites

  • Node.js ≥ 18
  • Python ≥ 3.9
  • A git repository to point at

Important: MemPalace Fork Required

⚠️ This MCP server requires a custom fork of MemPalace with multi-branch exclude-patterns support. The official pip install mempalace package will not work — you must use my fork (branch feat/exclude-patterns-config).

The easiest way to get this set up is the setup script (see Quick Start).


Quick Start

1. Run the setup script (recommended)

The fastest way to get up and running is the bootstrap script at scripts/setup-mempalace.sh. It handles the entire MemPalace setup:

  • Clones my fork with the required exclude-patterns support
  • Creates a Python virtual environment
  • Installs the fork in editable mode
  • Symlinks the mempalace CLI to ~/.local/bin
  • Prepares the palace data directory
  • Optionally runs the initial codebase index
# From the repo root:
./scripts/setup-mempalace.sh

# To also index your codebase right away:
./scripts/setup-mempalace.sh --mine

The script is fully configurable via environment variables:

Env var Default Description
MEMPALACE_FORK_URL https://github.com/2loch-ness6/mempalace My fork with required exclude-patterns support
MEMPALACE_FORK_BRANCH feat/exclude-patterns-config Branch containing the feature
MEMPALACE_FORK_DIR $HOME/.mempalace-fork Where to clone the fork
MEMPALACE_VENV_DIR $HOME/.mempalace Python venv location
MCP_PALACE_DIR <repo-root>/.palace/active Palace data directory
MCP_PALACE_WING code Palace wing (namespace)
MCP_REPO_DIR <repo-root> Source code to mine

After the script finishes, check that everything works:

mempalace --version
mempalace --palace ~/.mempalace/active status

2. Install the MCP server

npm install -g @mempalace/mcp-dev

Or use without installing via npx:

npx @mempalace/mcp-dev

3. Configure environment variables

export MCP_REPO_DIR=/path/to/your/repo        # defaults to process.cwd()
export MCP_PALACE_DIR=~/.mempalace/active      # MemPalace data directory
export MCP_PALACE_WING=code                    # wing name (namespace) for your repo
export MEMPALACE_PYTHON=/path/to/venv/bin/python3  # python with fork installed

If you used the setup script, set MEMPALACE_PYTHON to $HOME/.mempalace/bin/python3 (or wherever MEMPALACE_VENV_DIR points).

4. Add to your MCP client config

For Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "mempalace-mcp-dev": {
      "command": "npx",
      "args": ["@mempalace/mcp-dev"],
      "env": {
        "MCP_REPO_DIR": "/path/to/your/repo",
        "MCP_PALACE_DIR": "/home/you/.mempalace/active",
        "MCP_PALACE_WING": "code",
        "MEMPALACE_PYTHON": "/home/you/.mempalace/bin/python3"
      }
    }
  }
}

For Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "mempalace-mcp-dev": {
      "command": "npx",
      "args": ["@mempalace/mcp-dev"],
      "env": {
        "MCP_REPO_DIR": "/path/to/your/repo",
        "MCP_PALACE_DIR": "/home/you/.mempalace/active",
        "MCP_PALACE_WING": "code",
        "MEMPALACE_PYTHON": "python3"
      }
    }
  }
}

Environment Variables

Variable Default Description
MCP_REPO_DIR process.cwd() Absolute path to your git repository root
MCP_PALACE_DIR ~/.mempalace/active MemPalace data directory
MCP_PALACE_WING code Wing (namespace) name within the palace
MCP_LEDGER_PATH <repo>/.mcp-dev/ledger.jsonl Path to the change ledger file
MCP_TRACKS_DIR <repo>/conductor/tracks Directory containing project track folders
MEMPALACE_PYTHON python3 Python binary (must have the MemPalace fork installed)

MCP Tools Reference

Code Search

Tool Description
search_code Semantic search over indexed codebase. Accepts query (string) and optional scope (room name) and limit (int, default 10)
mine_changed_files Re-index files changed since the last git commit. Call after editing to keep search current

Branch & Git

Tool Description
get_branch_context Returns current branch, SHA, short SHA, and last 5 commit messages
get_changed_files Files changed vs HEAD~1 (or a specified base ref) with add/modify/delete status
get_git_diff Full diff output. Optional base ref and filePath for scoping

Change Ledger

Tool Description
record_change Append an entry to the ledger (files affected, description, reason, tags, optional track link)
read_ledger Read recent ledger entries. Optional limit (default 20) and branch filter

Dev Lifecycle

Tool Description
run_typecheck Run tsc --noEmit in a service directory. Returns structured errors with file/line/col/code
run_tests Run Jest in a service directory. Returns pass/fail counts and structured failure details

Project Tracks (Conductor)

Tool Description
list_tracks List all live tracks (directories containing plan.md)
get_track_plan Read plan.md for a track
update_track_plan Overwrite plan.md for a track
get_track_spec Read spec.md for a track
update_track_spec Overwrite spec.md for a track
get_track_index Read index.md for a track
update_track_index Overwrite index.md for a track
create_track Scaffold a new track directory with all four files

Project Tracks Methodology (Conductor)

Project Tracks is a lightweight, file-based feature management system. Each track is a directory containing up to four files:

conductor/tracks/
└── my-feature-20260501/
    ├── plan.md       ← living task list; the status source of truth
    ├── spec.md       ← technical spec, design decisions, acceptance criteria
    ├── index.md      ← short summary and linked resources
    └── metadata.json ← machine-readable metadata (id, owner, status, dates)

Rules:

  • A directory is a live track only if it contains plan.md.
  • plan.md is always read/written live from disk (never via MemPalace — no staleness).
  • Track slugs must match [a-zA-Z0-9][a-zA-Z0-9_-]* (max 100 chars).

Bootstrapping tracks in a new project:

# Create the tracks directory
mkdir -p conductor/tracks

# Use the MCP tool to scaffold a track
# (or copy from templates/track/ in this package)

Template files are included in templates/track/ for manual scaffolding.


MemPalace Configuration (mempalace.yaml)

Place a mempalace.yaml at the root of MCP_REPO_DIR to control how files are indexed. A generic example is provided in mempalace.yaml.example.

Key sections:

  • wing — must match MCP_PALACE_WING
  • exclude_patterns — gitignore-syntax patterns to skip during mining
  • rooms — semantic routing rules (keyword → room name) for search scoping

Development

git clone https://github.com/your-org/mempalace-mcp-dev
cd mempalace-mcp-dev
npm install
npm run build
npm test

See CONTRIBUTING.md for contribution guidelines.


License

MIT — see LICENSE.

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选