MCP Reference - OtherFunc

Overview

MCP (Model Context Protocol) lets AI assistants like Claude interact with OtherFunc directly — creating, managing, and executing functions without the user ever touching esoteric code. The OtherFunc MCP server exposes 8 tools that map to the OtherFunc API.

Setup

Prerequisites: an OtherFunc API key. Sign in on the home page to get one.

Install the MCP server binary:

cargo install --path crates/otherfunc-mcp

Set the OTHERFUNC_API_KEY environment variable (required). Optionally set OTHERFUNC_API_URL to point at a different API endpoint (defaults to https://api.otherfunc.com).

Configuration — Claude Desktop

Add the following to your claude_desktop_config.json:

{
  "mcpServers": {
    "otherfunc": {
      "command": "otherfunc-mcp",
      "env": { "OTHERFUNC_API_KEY": "of_your_key_here" }
    }
  }
}

Configuration — Claude Code

claude mcp add otherfunc otherfunc-mcp -e OTHERFUNC_API_KEY=of_your_key_here

Tools

Tool	Description	Parameters
`list_languages`	List available languages	(none)
`execute_code`	Run code without saving	`language`, `code`, `input?`
`execute_function`	Run a saved function	`id`, `input?`, `version?`
`create_function`	Create and deploy a function	`slug`, `language`, `name`, `code`, `description?`, `is_published?`
`get_function`	Get details and source code	`id`
`list_functions`	List your functions	(none)
`update_function`	Update metadata or code	`id`, `name?`, `description?`, `code?`, `is_published?`
`delete_function`	Permanently delete a function	`id`

Parameter Details

Tool	Parameter	Type	Required	Description
`execute_code`	`language`	string	yes	Language to execute (`brainfuck`, `forth`, `apl`, `lisp`)
`execute_code`	`code`	string	yes	Source code to run
`execute_code`	`input`	string	no	Stdin input for the program
`execute_function`	`id`	string	yes	Function ID (UUID)
`execute_function`	`input`	string	no	Stdin input for the function
`execute_function`	`version`	integer	no	Specific version to run (defaults to latest)
`create_function`	`slug`	string	yes	URL slug for the public endpoint
`create_function`	`language`	string	yes	Language for the function
`create_function`	`name`	string	yes	Human-readable name
`create_function`	`code`	string	yes	Source code
`create_function`	`description`	string	no	Description of what the function does
`create_function`	`is_published`	boolean	no	Whether the function is publicly accessible (default false)
`get_function`	`id`	string	yes	Function ID (UUID)
`update_function`	`id`	string	yes	Function ID (UUID)
`update_function`	`name`	string	no	New name
`update_function`	`description`	string	no	New description
`update_function`	`code`	string	no	New source code (creates a new version)
`update_function`	`is_published`	boolean	no	Update publish status
`delete_function`	`id`	string	yes	Function ID (UUID)

Example Workflow

A typical interaction where an AI assistant creates a function on your behalf:

User: “Make me an endpoint that reverses a string”
AI calls list_languages → picks Brainfuck
AI writes an implementation, calls execute_code to test it
AI calls create_function to deploy it with slug reverse-string
AI calls execute_function to verify the saved version works
AI returns the public URL: https://api.otherfunc.com/fn/reverse-string