TL;DR
TL;DR: MCPGod is an npm-installed CLI that gives you fine-grained control over Model Context Protocol servers — add, run, list, remove servers across clients, restrict specific tools per client, and inspect or call any tool directly from the terminal.
Source and Accuracy Notes
- Project page: npmjs.com/package/mcpgod
- Source repository: github.com/mcpgod/cli
- License: MIT (verified via npm package metadata)
- HN launch thread: news.ycombinator.com/item?id=43331973
What Is MCPGod?
MCPGod is a cross-platform CLI (oclif-powered) for managing MCP servers on Windows, macOS, and Linux. The core problem it solves: MCP servers are powerful but configuring them across different AI clients (Claude Desktop, Cursor, Windsurf, Cline) involves editing JSON config files manually. MCPGod replaces that friction with a terminal interface.
The README describes its philosophy simply:
MCPGod is a CLI tool designed to help developers manage MCP servers with speed and ease. Whether you need to add, run, list, or remove servers — or even interact with server tools — MCPGod provides a streamlined interface to handle all these tasks.
Setup Workflow
Step 1: Install
npm install -g mcpgod
Verify the installation:
mcpgod --version
Alternatively, run without installing:
npx -y mcpgod
Step 2: Add an MCP Server to a Client
mcpgod add @modelcontextprotocol/server-everything -c claude
This registers the server in Claude’s MCP configuration. To add only specific tools from a server:
mcpgod add @modelcontextprotocol/server-everything -c claude --tools=echo,add
Step 3: List Configured Servers
mcpgod list -c claude
Prints every MCP server currently registered for the specified client.
Step 4: Remove a Server
mcpgod remove @modelcontextprotocol/server-everything -c claude
Step 5: Run a Server Directly
mcpgod run @modelcontextprotocol/server-everything
Also supports local paths:
mcpgod run ./mcp-server.py
mcpgod run ./mcp-server-node.mjs
Step 6: Inspect and Call Tools
List every tool available on a server:
mcpgod tools @modelcontextprotocol/server-everything
Call a specific tool with parameters:
mcpgod tool @modelcontextprotocol/server-everything add a=59 b=40
mcpgod tool ./mcp-server.py echo message=hi
Deeper Analysis
Why MCPGod Matters
MCP is gaining adoption as a standard way for AI assistants to extend their capabilities via external tools. However, server management across clients is fragmented — each client has its own config format and no unified tooling. MCPGod fills that gap as a client-agnostic CLI layer.
Permission Control
The --tools filter on mcpgod add and the tool-level allow/block system give developers control over what a client can do with a given server. This matters in team environments where you may want Claude Desktop to access a filesystem server but not a network tool, while Cursor gets different permissions.
Logging
MCPGod logs every server run with timestamps and structured output, making it straightforward to debug what tools were called and when — something that is otherwise invisible in AI client UIs.
Practical Evaluation Checklist
- Install via npm and verify with
--version - Add a public MCP server to a local client config
- Use
--toolsfilter to restrict which tools are registered - Run
mcpgod listto confirm registration - Call a tool directly via
mcpgod toolwith key-value arguments - Review logged output for timestamped invocation records
Security Notes
- The CLI requires write access to your AI client’s MCP config directory — review what it writes before running with elevated privileges
--toolsfiltering is the primary security control; use it to limit exposure from untrusted servers- Tool calling via
mcpgod toolexecutes the underlying server logic locally — treat server arguments as you would any local command input
FAQ
Q: Which AI clients does MCPGod support?
A: MCPGod is client-agnostic in design. The -c flag accepts any client name, and the underlying config path is derived from your local MCP client setup. It has been tested with Claude Desktop, Cursor, Windsurf, Cline, and ChatGPT.
Q: Does MCPGod run servers for you?
A: mcpgod run starts a server process locally. For persistent server hosting, you still need a running server binary or Node.js process — mcpgod run provides the wrapper and logging, not the server itself.
Q: How is this different from the MCP SDK’s built-in tooling? A: The MCP SDK gives you the building blocks to create servers. MCPGod is the operational layer — adding, removing, listing, and invoking servers after they are built, without touching JSON config files by hand.
Conclusion
MCPGod turns MCP server management from a config-file editing task into a clean terminal workflow. For developers running multiple MCP servers across different AI clients, it removes the friction of manual JSON editing and adds tool-level permission controls. Install it with one npm command and it immediately works with any MCP-compliant server.
npm install -g mcpgod
Related Posts
dev-tools
Automotive Skills Suite for AI Engineering
Evaluate Automotive Skills Suite for APQP, ASPICE, HARA, safety-plan, and DIA workflows with setup notes, governance risks, and SME review guidance.
5/28/2026
dev-tools
awesome-agentic-ai-zh Roadmap Guide
Explore awesome-agentic-ai-zh as a Chinese agentic AI learning roadmap, with setup notes, track selection, study workflow, and evaluation guidance.
5/28/2026
dev-tools
Baguette iOS Simulator Automation Guide
Set up Baguette for iOS Simulator automation, web dashboards, device farms, gesture input, streaming, and camera testing with Xcode caveats.
5/28/2026