LLM Functions

v1.0.0Coding Agentsstable

Easily create LLM tools and agents using plain Bash/JavaScript/Python functions.

aiai-agentsai-toolsfunction-callingllm
Share:
751
Stars
0
Downloads
0
Weekly
0/5

What is LLM Functions?

LLM Functions is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to easily create llm tools and agents using plain bash/javascript/python functions.

Easily create LLM tools and agents using plain Bash/JavaScript/Python functions.

This server falls under the Coding Agents category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • Easily create LLM tools and agents using plain Bash/JavaScri

Use Cases

Create LLM tools with plain code
Build function-calling agents
Multi-language function definitions
sigoden

Maintainer

LicenseMIT
Languageshell
Versionv1.0.0
UpdatedMay 19, 2026
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux

Installation

Manual Installation

npx llm-functions

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use LLM Functions

LLM Functions is a framework that lets you build LLM tools and agents using plain Bash, JavaScript, or Python functions without any SDK boilerplate. It integrates with the AIChat CLI to expose your custom functions as callable tools, enabling tasks like weather lookups, command execution, web search, and code running. Developers use it to rapidly prototype function-calling agents by writing familiar scripts rather than learning a new abstraction layer.

Prerequisites

  • AIChat CLI installed (the only supported client for llm-functions)
  • argc framework installed (used to build and manage functions)
  • jq JSON processor installed
  • Git to clone the repository
  • An MCP-compatible AI client (Claude Desktop, etc.) or AIChat for local use
1

Clone the repository

Clone the llm-functions repository from GitHub to get the framework and example tools.

git clone https://github.com/sigoden/llm-functions
cd llm-functions
2

Install dependencies

Install the argc framework (used to build functions and manage tool definitions) and jq for JSON processing. On macOS with Homebrew: brew install argc jq. On Linux, follow the argc installation docs.

brew install argc jq
3

Build the functions

Run the build command to compile all function definitions into the required JSON format that the LLM can consume.

argc build
4

Check available tools

Verify that tools are correctly built and available. This lists all enabled tools from tools.txt.

argc check
5

Link to AIChat

Link the built functions to your AIChat installation so the CLI can find and invoke them. This sets up the AICHAT_FUNCTIONS_DIR environment variable linkage.

argc link-to-aichat
6

Enable tools in tools.txt

Edit tools.txt to list the function filenames you want enabled (one per line). For example, add get_current_weather.sh, execute_command.sh, or your own custom scripts.

7

Run with AIChat

Use AIChat in functions role to invoke tools, or run an agent by name. The framework handles routing the LLM's tool calls to your scripts.

aichat --role %functions% what is the weather in Paris?
aichat --agent todo list all my todos

LLM Functions Examples

Client configuration

Configure Claude Desktop to use llm-functions via npx as an MCP server. The server exposes your custom Bash/JS/Python functions as callable tools.

{
  "mcpServers": {
    "llm-functions": {
      "command": "npx",
      "args": ["llm-functions"],
      "env": {
        "AICHAT_FUNCTIONS_DIR": "/path/to/llm-functions"
      }
    }
  }
}

Prompts to try

Example prompts that exercise the built-in tools for weather, command execution, and web search.

- "What is the current weather in Tokyo?"
- "Run the command: ls -la ~/Documents"
- "Search the web for the latest news about AI agents"
- "Execute this Python snippet: print(sum(range(100)))"
- "List all my todos and mark the first one complete"

Troubleshooting LLM Functions

argc command not found after installation

Ensure argc is in your PATH. On macOS run 'brew install argc' and restart your terminal. On Linux, download the binary from https://github.com/sigoden/argc and place it in /usr/local/bin.

AIChat cannot find the functions directory

Run 'argc link-to-aichat' from the llm-functions directory, or manually set the AICHAT_FUNCTIONS_DIR environment variable to the full path of your llm-functions folder.

Tool not appearing even though it's in tools.txt

Run 'argc build' again after editing tools.txt — the build step regenerates functions.json which is what the LLM actually reads. Check 'argc check' output for any syntax errors in your function file.

Frequently Asked Questions about LLM Functions

What is LLM Functions?

LLM Functions is a Model Context Protocol (MCP) server that easily create llm tools and agents using plain bash/javascript/python functions. It connects AI assistants to external tools and data sources through a standardized interface.

How do I install LLM Functions?

Follow the installation instructions on the LLM Functions GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.

Which AI clients work with LLM Functions?

LLM Functions works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is LLM Functions free to use?

Yes, LLM Functions is open source and available under the MIT license. You can use it freely in both personal and commercial projects.

Browse More Coding Agents MCP Servers

Explore all coding agents servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Quick Config Preview

{ "mcpServers": { "llm-functions": { "command": "npx", "args": ["-y", "llm-functions"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use LLM Functions?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides