Open Interpreter
Open-source coding agent for the terminal that works with low-cost cloud models or locally via Ollama and LM Studio.
What it does
Open Interpreter is a coding agent for the terminal. You start it in a project folder and describe a task. The agent reads files, shows changes as diffs and runs commands. It builds on OpenAI’s Codex but isn’t tied to one provider: with /model you can switch to Kimi K3, DeepSeek, GLM, any OpenAI-compatible service, or local models via Ollama and LM Studio. With /harness it can emulate other agents such as Claude Code. It supports MCP, skills, hooks and AGENTS.md.
Who it suits
Development teams that want a coding agent running on low-cost or open models instead of committing to one large provider. It also suits companies that may only send source code to models they host themselves.
Data protection for businesses
The program runs on your machine and stores settings and sessions locally. By default, however, it sends usage statistics to Open Interpreter in the US; you can turn this off in the configuration. Where prompts and code go depends on the model you choose: with Ollama or LM Studio they stay on your own network. With cloud providers, their rules on server location, training and data processing apply. Open Interpreter does not publish a data processing agreement; offers for companies are available on request. For the paid desktop app and hosted models, the privacy policy says requests are logged for up to 30 days but not used for training. Commands run in a sandbox, and approval levels let you decide when the agent has to ask first.
Limits
The command-line program itself has no graphical interface, and setting it up takes technical knowledge. How good the results are depends heavily on the model. Without the sandbox, the agent can change or delete files. The earlier Python version is now maintained by the community.