process-mcp
An MCP server that allows AI assistants to manage background processes, enabling start, stop, monitoring, and querying of long-running shell commands without blocking the conversation.
README
process-mcp
A local MCP (Model Context Protocol) server for managing background processes.
process-mcp lets AI assistants and other MCP clients start, stop, monitor, and query long-running shell commands without blocking the conversation. Instead of waiting for a command to finish, clients can start it in the background and receive a notification when it exits.
Features
- Start background processes with custom arguments, working directory, and environment variables
- Stop processes gracefully with SIGTERM or forcefully with SIGKILL
- Query real-time status and retrieve recent log output
- List all managed processes and resolve their system PIDs
- Receive
process/exitnotifications when a background process terminates - Short, friendly process IDs via
nanoid(6 characters) - Strict input validation with
zod - Runs over stdio — no HTTP or SSE ports required
Installation
Requires Bun.
bun install
bun run build
To run the server directly during development:
bun run dev
To run the compiled server:
bun run start
Configuration
Add process-mcp to your MCP client configuration. For opencode, edit ~/.config/opencode/opencode.json:
{
"mcpServers": {
"process-mcp": {
"command": "bun",
"args": [
"run",
"/path/to/process-mcp/dist/index.js"
]
}
}
}
Adjust the absolute path to match your local clone location.
Tools Reference
start
Start a background process.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
command |
string | yes | Executable or command to run |
args |
string[] | no | Arguments passed to the command |
cwd |
string | no | Working directory for the process |
env |
record | no | Extra environment variables |
notifyOnExit |
boolean | no | Send a process/exit notification when the process ends |
Example response:
{
"id": "a1b2c3",
"pid": 12345,
"status": "running"
}
stop
Stop a managed process.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
id |
string | yes | Process ID returned by start |
force |
boolean | no | Use SIGKILL instead of SIGTERM (default: false) |
Example response:
{
"id": "a1b2c3",
"pid": 12345,
"status": "stopped"
}
get-status
Get the current status of a process.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
id |
string | yes | Process ID |
Returns: "running" | "stopped" | "exited" | "crashed" | "not-exists"
get-log
Retrieve the stdout/stderr log of a process.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
id |
string | yes | Process ID |
lastLines |
number | no | Number of recent log lines to return |
Example response:
{
"lines": [
"Build started...",
"Compiled 42 files",
"Build finished"
]
}
list
List all managed processes.
Parameters: none
Example response:
{
"processes": [
{ "id": "a1b2c3", "pid": 12345, "status": "running" },
{ "id": "d4e5f6", "pid": null, "status": "exited" }
]
}
get-pid
Get the system PID of a managed process.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
id |
string | yes | Process ID |
Returns: PID number or null
Notification
When a process is started with notifyOnExit: true, the server sends a process/exit notification once the process terminates. The notification payload includes the process ID and final status.
This allows clients such as opencode to resume a conversation after a long-running task completes, without polling for status.
Development
Run the test suite:
bun test
Build the project:
bun run build
Run the server from source:
bun run dev
License
MIT
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。