ShellifyAI

Secure shell execution for AI agents

// works with Claude, OpenAI, and more

Secure Shell Execution & AI Agent Orchestration.

ShellifyAI lets your AI agents execute real shell commands in secure, sandboxed environments—or delegate complex tasks to AI agents (OpenAI Codex, Claude Code) that autonomously write code, install packages, and iterate until complete.

Add shell execution and agent orchestration to your AI app in minutes. Define the tool, and Shellify handles the rest—sandboxing, security, file artifacts, and streaming. Get Started

Claude

OpenAI

POST /v1/execute

$ curl -X POST -H "x-api-key: $API_KEY" ...

api/chat/route.ts

import { streamText, tool } from "ai";
import { z } from "zod";
 
// Define the tool - Shellify handles execution
const result = streamText({
  model: openai("gpt-5.1"),
  tools: {
    local_shell: tool({
      description: "Execute shell commands",
      parameters: z.object({
        command: z.string()
      }),
      execute: async ({ command }) => {
        // Forward to Shellify API
        return fetch(`https://shellifyai.com/v1/execute`, {
          method: "POST",
          body: JSON.stringify({ command })
        }).then(r => r.json());
      }
    })
  }
});

Features

Long-running processes

Safely execute tasks that require extended computation or monitoring without risking system stability or security.

Real time streaming

Listen to standard output & error streams live when executing long-running commands.

Multi-adapter support

Run direct sandbox execution as a code interpreter, or delegate complex tasks to AI agents like Claude Code and OpenAI Codex.

Session persistence

Reuse sessions for file persistence across commands. Files created in one command are available in the next.

File artifacts

Automatic file artifact detection and upload. Generated files are stored and accessible via URLs.

Secure sandboxing

Execute commands in isolated environments with automatic cleanup and timeout management.

What Shellify handles for you:

✓ Sandboxed execution
✓ Security isolation
✓ File artifact uploads
✓ Streaming output
✓ Session persistence
✓ Timeout management

Your
App

Shellify

Define tool → Forward to API → We execute securely

Examples

Real code examples from our documentation. Copy and paste to get started.

example.ts

import OpenAI from "openai";

const client = new OpenAI();

// 1. Define the tool for OpenAI
const tools = [{
  type: "function",
  name: "local_shell",
  description: "Execute shell commands in a secure sandbox",
  parameters: {
    type: "object",
    properties: {
      command: { type: "string", description: "Shell command to execute" }
    },
    required: ["command"]
  }
}];

// 2. Create your request
const response = await client.responses.create({
  model: "gpt-5.1",
  input: "Create a Python file that prints Hello World and run it",
  tools
});

// 3. When model calls the tool, forward to Shellify
for (const item of response.output) {
  if (item.type === "function_call" && item.name === "local_shell") {
    const result = await fetch(
      `https://shellifyai.com/v1/execute`,
      {
        method: "POST",
        headers: {
          "Content-Type": "application/json",
          "x-api-key": process.env.SHELLIFYAI_API_KEY!
        },
        const args = JSON.parse(item.arguments);
        body: JSON.stringify({
          adapterType: "local_shell",
          tool: "local_shell",
          payload: { command: args.command, intent: args.intent }
        })
      }
    ).then(r => r.json());

    // 4. Return result to model
    await client.responses.submitToolOutputs(response.id, {
      tool_outputs: [{ tool_call_id: item.call_id, output: JSON.stringify(result) }]
    });
  }
}

Get Started

Create a project and get your API key in seconds.

npm i @shellifyai/shell-tool