Open Interpreter Open source

What it does

Open Interpreter is a terminal-based coding agent, rewritten in Rust as a fork of OpenAI's Codex harness, aimed at getting strong agentic performance out of budget-friendly open-weight models (the README calls out Kimi K3 and GLM 5.3) instead of only frontier proprietary ones. It executes commands in native sandboxes and can drive both terminal and GUI applications.

What is inside

• Multiple harness "personalities" (native, claude-code-style, kimi-code-style, etc.) so the agent's behavior can emulate other popular CLIs
• A one-line compatibility path for OpenAI's Codex SDK
• Computer-use support via agent-browser and trycua integration for controlling web and native apps
• Support for MCP servers, skills, hooks and fine-grained permission management
• An Agent Client Protocol mode so it can be embedded as an agent inside compatible editors

Works with

Designed to plug into multiple LLM providers/models rather than one vendor, with explicit emphasis on open models (Kimi K3, GLM 5.3) alongside OpenAI-compatible endpoints.

How to install or connect

curl -fsSL https://www.openinterpreter.com/install | sh

Windows:

irm https://www.openinterpreter.com/install.ps1 | iex

Maintenance and safety

Apache-2.0 licensed with a large, active history (68k+ stars, 5.9k forks, 10,000+ commits). Note this is a substantial Rust rewrite quite different in architecture from the classic Python "Open Interpreter" concept, built around sandboxed command execution and computer-use — review its sandboxing and permission settings before granting it broad system or browser control, as with any agent that executes shell commands and drives applications.

Who should use it

Developers who want a terminal coding agent optimized for cost-efficient open-weight models, or who want one harness that can emulate the UX of several other popular CLI agents. It is free and self-hosted (you bring your own API keys or local model endpoint), so teams that are already paying per-token for a frontier-model CLI may find it a cheaper daily driver for routine tasks, reserving pricier models for the hardest problems. Because it is still a young rewrite of a previously very different project, expect some rough edges and configuration required compared to more mature single-vendor CLIs.