mcp-cli: Claude Code Skill & Agent Tool
An open-source skill by github (opens in a new tab), published in the github/awesome-copilot (opens in a new tab) repository on GitHub. A command-line bridge to MCP servers: your agent lists the servers you have, inspects their tools and calls them, so external APIs and data sources are one command away.
Install mcp-cli
Requires Node.jsnpx skills add github/awesome-copilot --skill mcp-cliRun this command in your project terminal to install this skill for Claude Code.
What this skill does
mcp-cli gives your agent a command-line interface to Model Context Protocol servers. MCP is how agents reach external systems — GitHub, file systems, databases and APIs — and this skill lets the agent use any server you have configured through a handful of simple commands instead of a dedicated integration.
The workflow has four steps. Discover lists every configured server and its tool names. Explore shows a server's tools with their parameters, and one flag adds their descriptions. Inspect returns the full JSON input schema of a single tool, so the agent knows exactly what to send. Execute calls the tool with JSON arguments, passed inline, through a heredoc when the content contains quotes, or piped from a file or another command. A grep command searches tools by name across all servers.
For scripting, output can be returned as JSON or as raw text, and distinct exit codes separate client errors, server errors and network failures, so the agent can react correctly when something goes wrong. Commands chain with standard shell tools such as jq. Use it with Claude Code to try an MCP server quickly, automate calls in scripts, or debug a tool's inputs and outputs.
Key Capabilities
- Lists your configured MCP servers and their tools
- Shows tool parameters, descriptions and full JSON input schemas
- Calls tools with JSON arguments inline, by heredoc or from a pipe
- Searches tools by name across every server
- JSON or raw output and distinct exit codes for scripting
How to use with Claude Code & AI Agents
Install the skill
Run the command above at the root of your project. The CLI asks which agents to set up — Claude Code, Cursor and other supported agents — and adds mcp-cli to the project, so teammates who pull the repository get it too. Add-gto install it for your user account across every project.Open a new agent session
Start a fresh Claude Code session in that project so the agent loads the new skill. Have ready: The mcp-cli command, with MCP servers configured.Ask in plain words
Describe the task, for example:“List my MCP servers and use the filesystem server to read README.md.”The agent matches your request to the skill's description and follows its SKILL.md instructions step by step. Read them on GitHub before using the skill on a system that matters.