cli-use

Overview

Cli-use transforms any MCP server into a native command-line interface through a single command, dramatically reducing token consumption for AI agents. It eliminates schema discovery, request framing, and response parsing overhead by converting verbose JSON-RPC calls into terse CLI commands. The tool supports both one-shot execution and daemon mode for improved performance, with features like call caching, batch/pipe mode, shell completions, and OpenAPI export. Built with pure Python standard library and maintains compatibility across Linux, macOS, and Windows.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• addadd <alias> <path>

  Install and register an MCP server as a CLI alias

  Try:Add the filesystem MCP serverTry:Install git MCP serverTry:Register new server
• list

  List available MCP servers in the built-in registry

  Try:Show available MCP serversTry:List all registered servers
• daemon

  Manage MCP server daemons for faster repeated calls

  Try:Start filesystem daemonTry:List running daemons
• openapi

  Generate OpenAPI spec from cached MCP server tools

  Try:Generate OpenAPI spec for filesystem toolsTry:Export API documentation
• completions

  Generate shell completion scripts for MCP tools

  Try:Generate bash completionsTry:Setup shell autocompletion
• batch

  Run JSON pipeline across multiple MCP server aliases

  Try:Process batch operationsTry:Run pipeline across servers
• run

  Run a specific MCP tool directly from command line

  Try:Run greet tool with argumentsTry:Execute tool directly
• mcp-list

  List tools available from an MCP server

  Try:List available tools in serverTry:Show MCP server capabilities
• convert

  Convert MCP server to standalone CLI script

  Try:Convert server to CLI scriptTry:Create standalone tool
• fs

  Filesystem operations: read, write, list, and search files

  Try:List directory contentsTry:Read text fileTry:Search files by pattern
• gh

  GitHub operations: issues, PRs, and repos

  Try:Create GitHub issueTry:List pull requestsTry:Get repository info
• sqlite

  Query and inspect SQLite databases

  Try:Execute SQL queryTry:List database tables

Comparable tools

Installation

git clone https://github.com/cli-use/cli-use.git
cd cli-use
pip install -e .

{
  "mcpServers": {
    "cli-use": {
      "command": "python",
      "args": ["-m", "cli_use.server"]
    }
  }
}

FAQ

How much token reduction does cli-use provide?

Measured tests show 60-80% token reduction across different session sizes, with significant savings in schema discovery (80% cheaper) and per-call I/O (59% cheaper).

What MCP servers are pre-configured?

Cli-use comes with built-in support for 10 servers: fs (filesystem), memory (knowledge graph), gh (GitHub), git, sqlite, time, fetch, puppeteer, brave (web search), and slack.

Compare cli-use with