Smoke Monkey Harness: A Zero-Dependency TypeScript Runtime for Autonomous AI Agents

Overview

Smoke Monkey Harness is an open-source tool designed to turn Large Language Models into autonomous engineering agents. It is built with TypeScript and works without needing any extra runtime libraries, relying only on the built-in tools that come with Node.js. This makes it lightweight, secure, and easy to integrate into existing projects. It is released under the MIT license, meaning developers can use it freely for personal or commercial work.

Benefits

Smoke Monkey Harness offers several key advantages for developers and engineering teams. First, it has zero runtime dependencies. This means it does not require installing heavy external packages or databases, which reduces supply chain risks and keeps the system lean. Second, it uses a deterministic six-phase state machine to manage tasks. This approach replaces fragile loops with a clear process that includes exploring the workspace, planning, editing, verifying, recovering from errors, and completing the task. This ensures safe and predictable execution. Third, it includes a human-in-the-loop feature. Users can set permission policies that pause the agent for approval before it performs sensitive actions like deleting files or running commands. Fourth, it manages context automatically. When conversation limits are reached, the tool summarizes history to allow sessions to continue indefinitely. Finally, it supports resumable sessions. Work can be paused and resumed later using stable session files, which is helpful for long-running tasks.

Use Cases

Smoke Monkey Harness is ideal for developers who want to build autonomous AI agents for software engineering tasks. It can be used directly within a TypeScript codebase to automate tasks like refactoring code, writing tests, or managing Git repositories. The tool comes with 24 built-in tools for file management, terminal commands, searching, and Git operations. It also supports a system for loading engineering skills on demand, covering areas like frontend design, QA testing, backend API design, and DevOps security. Developers can embed the agent into their applications or use it as a standalone service. It also integrates with popular IDEs and agents through the Model Context Protocol, allowing users to scaffold projects and plan architectures directly from their editor. The tool supports multiple AI providers, including Claude, NVIDIA, Google, OpenAI, and local models like Ollama, making it versatile for different environments.

Pricing

Smoke Monkey Harness is completely free to use. It is open-source and released under the MIT license. There are no hidden costs, royalties, or subscription fees. Developers can use it in commercial, enterprise, or personal projects without any restrictions.

Vibes

The community and developers have responded positively to Smoke Monkey Harness. It is praised for its transparency and lack of hidden telemetry. Unlike some other tools, it does not send data to external servers or have artificial throttles. Users appreciate that all data stays within their private environment. The tool is seen as a robust foundation for building reliable engineering assistants. Its lightweight nature and support for multiple AI models make it a popular choice for teams looking for flexibility and control.

Additional Information

Smoke Monkey Harness is built to be provider-agnostic, meaning it works with any OpenAI-compatible endpoint. It is optimized for models like Claude Opus 5.5 and Claude Sonnet 5 but also supports NVIDIA Nemotron, Google Gemini, OpenAI GPT-4o, and local models via Ollama. The project includes an official Model Context Protocol server, which allows it to be driven by IDEs like Cursor and Codex. It also offers a companion chat interface package for a better user experience. The tool distinguishes itself from other frameworks by requiring no database and using native TypeScript instead of Python-heavy alternatives.