Open Interpreter
A terminal coding agent built on Codex, tuned for low-cost open models like Kimi K3, DeepSeek and Qwen
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Runs terminal commands in a sandbox, but with access to the project's files
Install
Manual install
curl -fsSL https://www.openinterpreter.com/install | shOn Windows: irm https://www.openinterpreter.com/install.ps1 | iex
This is third-party code. Review the repository files before installing.
What it does
Open Interpreter is a Rust rewrite of Codex that emulates different agent harnesses to get the most out of budget models. The /harness command switches on the fly between modes that mimic Claude Code, Kimi Code, Qwen Code, DeepSeek TUI and others. It runs as an ACP agent in compatible editors, supports MCP, hooks, and the shared .agents/skills directories, and can drive a browser or native apps through a built-in QA skill.
Who it is for. For developers who want a terminal coding agent on open or low-cost models without paying full price for Claude or GPT.
Good fit when
- You want a terminal coding agent but want to save money using DeepSeek, Kimi or Qwen
- You need to emulate a specific harness's behavior (Claude Code, Kimi Code) on top of a different model
- You already use the Codex SDK and want to swap the backend with a one-line override
Not a fit when
- You need official support and guarantees from Anthropic or OpenAI rather than a community fork
- The stock Codex CLI without third-party harness emulation is enough
Example request
Switch the harness to kimi-code and figure out why test_parser.py is failingLimitations
Paid providers like Kimi, DeepSeek or GLM need their own keys and paid API access. It is a Codex fork, not an independent build, so updates depend on how quickly the project syncs with upstream.
How to disable. Remove the interpreter binary and the ~/.openinterpreter config folder.
Security check
- Runs terminal commands in a sandbox, but with access to the project's files
README in short
The README presents Open Interpreter as a coding agent optimized for cheap models. The core idea is harness emulation: the /harness command makes the agent mimic native, claude-code, zcode, kimi-code, kimi-cli, qwen-code, deepseek-tui, swe-agent and others. The project is compatible with the ACP protocol and Codex exec, uses shared .agents/skills directories, and supports hooks, MCP and permissions. A built-in QA skill drives a browser via agent-browser and native apps via trycua. Apache-2.0 licensed.
FAQ
Is this the official Open Interpreter?
No, this is a new Rust version built on Codex by a separate team. The original Python project lives on as the endolith/open-interpreter community fork.
Does it work with Claude or GPT?
Yes, via provider selection in /model, but the project's main focus is low-cost open models.
Related
Open-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins
A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself
An open source coding agent for the terminal and desktop with build and plan modes
Open prompt library with a Claude Code plugin, MCP server and CLI: search, fetch and improve prompts and skills from an agent