Open Interpreter

Open Interpreter is an open-source terminal coding agent built for low-cost and open-weight models. It is a fork of Codex, so it inherits much of that setup.

You run it in a project folder, and it can read files, edit code, run commands and ask before it goes beyond its sandbox. It ships model-specific harnesses for models such as Kimi, Qwen and DeepSeek, so the agent loop suits each model. Existing AGENTS.md instructions, skills and MCP servers carry over, and slash commands such as /init, /skills and /mcp help you configure it.

• Model-specific harnesses for Kimi, Qwen and DeepSeek
• Sandboxed by default, with file and network access gated
• Asks approval before commands outside the sandbox
• Carries over AGENTS.md, skills and MCP setup
• Slash commands such as /init, /skills and /mcp

Developers who want a terminal agent that works well with open-weight and lower-cost models, and people who want a sandboxed agent by default.

The project is Apache 2.0 licensed and free to use. Model costs depend on the provider you connect, which can include hosted services or local models run through Ollama or LM Studio. The sandbox runs on macOS, Linux and Windows, but check the approvals you grant before running commands on important projects.