Mcp Server Commands
Expose shell and CLI execution to your agent through MCP when scaffolding projects, running tests, or invoking local tooling.
Overview
io.github.bytedance/mcp-server-commands is a MCP server for the Build phase that runs arbitrary shell commands with an optional working directory.
What is this MCP server?
- Runs arbitrary commands via MCP tools from @agent-infra/mcp-server-commands
- Optional cwd argument to pin the working directory per invocation
- stdio transport for local agent hosts
- Alternate SSE transport on http://127.0.0.1 with configurable port (default 8089)
- Same npm package supports both embedded stdio and networked SSE modes
- Server version 1.0.0
- Default SSE port 8089
- 2 transport modes: stdio and SSE
Community signal: 36.3k GitHub stars.
What problem does it solve?
Agents stall when every build, test, or git step requires you to run commands manually and paste stdout back into the chat.
Who is it for?
Experienced solo builders on trusted laptops who want maximum agent autonomy for scripts, package managers, and local dev CLIs.
Skip if: Shared production servers, junior workflows without sandboxing, or teams that prohibit shell access from AI tools.
What do I get? / Deliverables
After you register the server with cwd and transport settings, the agent can execute local commands through MCP and read results in the same session.
- MCP-accessible command execution with configurable working directory
- Choice of stdio or localhost SSE transport for the same package
- Faster local script and package-manager loops inside agent sessions
Recommended MCP Servers
Journey fit
Command execution is core agent tooling during Build when the product still lives on your machine and scripts are the fastest feedback loop. The canonical shelf is agent-tooling because the server’s job is extending the agent with process execution, not shipping features directly to users.
How it compares
High-power MCP shell bridge, not a curated task skill with fixed safe commands.
Common Questions / FAQ
Who is io.github.bytedance/mcp-server-commands for?
Advanced indie developers using MCP hosts who need the agent to run local CLI and shell workflows during active coding sessions.
When should I use io.github.bytedance/mcp-server-commands?
Use it during Build when scripted feedback is faster than manual terminal copy-paste and you accept the security tradeoffs of arbitrary execution.
How do I add io.github.bytedance/mcp-server-commands to my agent?
Install @agent-infra/mcp-server-commands, choose stdio or SSE with port, set cwd to your repo root in MCP config, and restart the agent host.