openinterpreter

Why we included this project

Open Interpreter is a fork of OpenAI's Codex, reimplemented in Rust to emulate the agent harnesses that get the best results from low-cost open models. Its headline feature is a /harness command that swaps between several harnesses, including native, Claude Code, Kimi Code, Qwen Code, and DeepSeek TUI, so you can match the prompting style to whatever model you are running. It runs commands under native sandboxing on macOS, Linux, and Windows, and it speaks the Codex exec protocol and the Agent Client Protocol, which means it can drop into editors and existing Codex SDK workflows with a one-line binary override. Teams already using shared AGENTS.md instructions, MCP, and .agents/skills directories will find it portable rather than locked into a proprietary format. A built-in QA skill even lets the agent drive real web and native apps for testing, so it reaches beyond code editing into interface verification.