devcontainer-mcp
A minimal MCP server that exposes a single bash tool to execute commands inside a devcontainer, returning real output and exit codes.
README
devcontainer-mcp
A minimal Model Context Protocol server that
exposes exactly one tool — bash — which mimics Claude Code's built-in
Bash tool but executes every command inside a devcontainer instead of on
the host.
The command runs via:
devcontainer exec --workspace-folder <folder> [--remote-env K=V ...] bash -lc "<command>"
and the tool returns the command's real combined stdout/stderr and exit code — not a static "Executed command …" string.
Addressed as
mcp__devcontainer__bash(server namedevcontainer, tool namebash).
Why
This is a Python rewrite of the TypeScript
mcp-devcontainers
server, trimmed to a single Bash-like tool and fixing its main defect: the
original returned a fixed string and tee'd output to a file, so the model never
saw what actually happened. Here, stdout/stderr and the exit status flow back in
the tool result, and a non-zero exit (or a timeout) flags the result as an
error.
It deliberately does not expose up, run-user-commands, or a generic
exec — bringing the container up is out of scope for v1.
Requirements
- Python ≥ 3.10
- The
@devcontainers/cliavailable asdevcontaineronPATH(or override viaDEVCONTAINER_CLI, e.g.npx @devcontainers/cli) - A running devcontainer for the target workspace (start it yourself with
devcontainer up --workspace-folder .)
Install / run
Run straight from the repo with uvx — no install
step:
uvx --from git+https://github.com/vovayartsev-dice/devcontainer-mcp devcontainer-mcp
The server speaks MCP over stdio.
.mcp.json
Add it to your project's .mcp.json (the server key is what produces the
mcp__devcontainer__* tool prefix):
{
"mcpServers": {
"devcontainer": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/vovayartsev-dice/devcontainer-mcp",
"devcontainer-mcp"
],
"env": {
"DEVCONTAINER_WORKSPACE_FOLDER": "${workspaceFolder}"
}
}
}
}
Configuration (environment variables)
| Variable | Default | Purpose |
|---|---|---|
DEVCONTAINER_CLI |
devcontainer |
CLI invocation; shell-split, so npx @devcontainers/cli works. |
DEVCONTAINER_WORKSPACE_FOLDER |
current dir | Default host workspace folder when the workspaceFolder arg is omitted. |
The bash tool
| Field | Type | Required | Default | Notes |
|---|---|---|---|---|
command |
string | ✅ | — | Run via bash -lc "<command>" inside the container. Pipes/&&/globs/redirects work. |
cwd |
string | workspace root | Working dir inside the container: cd <cwd> && ( <command> ). |
|
env |
object {KEY: VAL} or list KEY=VAL |
— | Forwarded as repeated --remote-env. |
|
timeout |
integer (ms) | 120000 (max 600000) |
On timeout the process group is killed; result flagged as error. | |
description |
string | — | Logged for observability; otherwise unused. | |
workspaceFolder |
string | cwd / env var | Host path holding the .devcontainer config. |
Result
- Content is the combined stdout + stderr (interleaved), truncated to
~30 000 chars with a
[output truncated]marker if longer. - On a non-zero exit, the exit code is appended and
isErroristrue. - On a timeout, a timeout note is appended and
isErroristrue.
Security / scope
- The command string is interpreted by bash only inside the container. The
host process is launched argv-only (never
shell=True, never a hostsh -c), so host-side shell injection isn't possible. - The container is the blast radius — this server does not mount the Docker
socket, add
--privileged, or expose any tool other thanbash. - v1 intentionally skips background execution / streaming and container bring-up.
Development
uv run devcontainer-mcp # run the server from a checkout
uv build # build wheel + sdist
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。