About this project
Open Interpreter is a coding agent designed for low-cost models, built as a fork of OpenAI's Codex with a focus on emulating agent harnesses that get the best performance from inexpensive models. It runs in the terminal with a text-based interface (type `i` or `interpreter` to start).
Key capabilities include harness emulation via the `/harness` command, with options such as native, claude-code, zcode, kimi-code, qwen-code, deepseek-tui, swe-agent, and minimal. It also supports switching providers and models from the TUI using `/model`, and can run any OpenAI-compatible provider through Chat Completions using `interpreter --chat-completions` or `interpreter exec --chat-completions`.
Open Interpreter is compatible with the Agent Client Protocol (ACP), allowing it to work in ACP-compatible editors and clients via `interpreter acp`. It also speaks the OpenAI Codex exec protocol, so existing Codex SDK integrations can use it with a one-line binary override. A local compatibility check is available via `scripts/test-codex-sdk-compat.sh`.
The project emphasizes portability: it prefers shared, tool-neutral standards and directories, including repository `AGENTS.md`, shared `.agents/skills` directories, MCP, ACP, and the Codex exec protocol. Product-specific configuration remains under `~/.openinterpreter` for runtime state that lacks a practical shared standard.
It ships with a QA skill for computer use, letting any model operate and test interfaces—driving web apps in a real browser via agent-browser, or operating and testing native apps via trycua.
Features include native sandboxing on macOS, Linux, and Windows; runtime inspection or switching of Rust-native model harnesses; web and native app testing through the built-in QA skill; running as an ACP agent for editors; reuse of shared `AGENTS.md` instructions and `.agents/skills` directories; and support for exec, MCP, skills, hooks, permissions, and `AGENTS.md`.
Documentation covers terminal usage, quickstart, installation, configuration, CLI reference, harnesses, model provider guides (including Kimi K3, DeepSeek, and Z.AI/GLM/ZCode), ACP, Codex SDK, portability, and sandbox/approvals.
The repository notes that this is the new Rust version of Open Interpreter based on Codex. The original Python project is maintained as a community fork at endolith/open-interpreter. Licensed under Apache-2.0.
Comments
0 Rating appears after 10 ratings
Sign in to join the discussion.