just-runner-mcp
Exposes Just recipes as MCP tools, allowing AI assistants to discover and execute project commands defined in Justfiles.
README
Just an MCP Command Runner
A Model Context Protocol (MCP) server that exposes Just recipes as MCP tools, allowing AI assistants to discover and execute project commands defined in Justfiles.
Quick Start
Install and run the MCP server with npx:
# Install just command runner if not already installed
# See https://github.com/casey/just#installation
# Run the MCP server pointing to a directory with a Justfile
npx just-runner-mcp --justfile ./path/to/your/justfile
# Or run from current directory (will search for Justfile)
npx just-runner-mcp
# Test via MCP inspector web UI
npx -y @modelcontextprotocol/inspector -- npx -y just-runner-mcp --justfile tests/support/basic.justfile
MCP Client Configuration
Add to your MCP client configuration (e.g., Claude Desktop):
{
"mcpServers": {
"just-runner": {
"command": "npx",
"args": ["just-runner-mcp", "--justfile", "/path/to/your/project"]
}
}
}
Why Just?
Just is a command runner that saves and runs project-specific commands in a justfile. It's an effective choice for exposing commands/scripts to AI assistants because:
- Simple syntax: Justfiles use a clean, make-inspired syntax that's easy to read and write
- Cross-platform: Works on Linux, macOS, Windows, and other Unix-like systems
- Language agnostic: Recipes can be written in any language (bash, python, etc.)
- Self-documenting: Recipes can include documentation comments that describe what they do
- Parameterized: Recipes can accept arguments, making them flexible and reusable
- Dependency management: Recipes can depend on other recipes, ensuring proper execution order
Tool Generation
The MCP server automatically parses Justfiles and generates MCP tools for each recipe. Each tool is named after the recipe and includes:
- Name: The recipe name from the justfile
- Description: Extracted from documentation comments (lines starting with
#above the recipe) - Parameters: Automatically detected recipe parameters with their default values and types
For example, this justfile recipe:
# Compile Latex Document. FILEPATH must be an absolute path
compile-latex FILEPATH:
tectonic {{FILEPATH}}
# Compile Typst Document. FILEPATH must be an absolute path
compile-typst FILEPATH:
typst compile --format pdf {{FILEPATH}}
typst compile --format png {{FILEPATH}} "page-{0p}-of-{t}.png"
Would expose the following tools to the MCP client:
{
"tools": [
{
"name": "compile-latex",
"description": "Run just recipe: compile-latex\nCompile Latex Document. FILEPATH must be an absolute path",
"inputSchema": {
"type": "object",
"properties": {
"FILEPATH": {
"type": "string",
"description": "Required"
}
},
"required": ["FILEPATH"]
}
},
{
"name": "compile-typst",
"description": "Run just recipe: compile-typst\nCompile Typst Document. FILEPATH must be an absolute path",
"inputSchema": {
"type": "object",
"properties": {
"FILEPATH": {
"type": "string",
"description": "Required"
}
},
"required": ["FILEPATH"]
}
}
]
}
CLI Options
The MCP server supports several command-line options:
Usage: just-mcp [options]
Expose Justfile recipes as MCP tools
Options:
--justfile <path> Path to Justfile (default: "Justfile")
--just-binary <path> Path to just binary (default: "just")
--timeout <ms> Timeout for recipe execution in milliseconds (default: 30000)
--list-tools Print available tools and exit
--help Show help message
-h display help for command
Examples
This section demonstrates a practical Justfile for a development project and the MCP tools it generates. Each recipe becomes a tool that AI assistants can discover and execute.
Simple Development Justfile
Here's a justfile for a typical development workflow:
# Install project dependencies
install:
npm install
# Start the development server
dev port="3000":
echo "Starting dev server on port {{port}}"
npm run dev -- --port {{port}}
# Run all tests
test:
npm test
# Deploy to staging environment
deploy-staging: build test
echo "Deploying to staging..."
./scripts/deploy.sh staging
MCP Tools Output
When you run just-runner-mcp --list-tools, you get the following JSON response showing all available tools:
{
"tools": [
{
"name": "install",
"description": "Run just recipe: install\nInstall project dependencies",
"inputSchema": {
"type": "object",
"properties": {},
"required": []
}
},
{
"name": "dev",
"description": "Run just recipe: dev\nStart the development server",
"inputSchema": {
"type": "object",
"properties": {
"port": {
"type": "string",
"description": "Parameter: port (default: 3000)"
}
},
"required": []
}
},
{
"name": "test",
"description": "Run just recipe: test\n",
"inputSchema": {
"type": "object",
"properties": {},
"required": []
}
},
{
"name": "deploy-staging",
"description": "Run just recipe: deploy-staging\nDeploy to staging environment",
"inputSchema": {
"type": "object",
"properties": {},
"required": []
}
}
]
}
With these tools available, an AI assistant can help you with common development tasks like "install the dependencies", "start the dev server on port 8080", "run the tests", or "deploy to staging".
Alternatives
There are other approaches to integrating Just with MCP:
- just-mcp - A Rust-based MCP server that provides more comprehensive Justfile support and handling. This cargo package offers additional features beyond just exposing recipes as tools to the LLM.
This repository (just-runner-mcp) takes a minimal approach, focusing specifically on exposing Just recipes as MCP tools for AI assistants to discover and execute.
Development
# Run all tests
just test
# Run specific test file
just test tests/integration.test.ts
License
This project is licensed under the MIT License - see the LICENSE file for details.
推荐服务器
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。
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。