Open Interpreter

Open Interpreter is an open-source terminal coding agent maintained by Killian Lucas and the Open Interpreter community. It is designed to act as an execution environment where language models can read files, write code, run shell commands, and interact with the local operating system. In daily use, Open Interpreter runs directly in your terminal via a conversational TUI. Developers interact with the agent across standard workflows such as debugging, refactoring, and automating shell tasks. It supports a broad range of models and providers, including local inference engines like Ollama and LM Studio, hosted APIs like OpenAI, Anthropic, OpenRouter, and Amazon Bedrock, as well as cost-effective Chinese and open-weight models from DeepSeek, Moonshot/Kimi, Z.ai, and Alibaba/Qwen. What distinguishes Open Interpreter is its focus on model-specific harness emulation and robust OS-level sandboxing. Instead of treating all models with a single generic prompting strategy, it shapes the execution loop and request structure to fit the specific strengths of models like DeepSeek, Qwen, or Kimi. Furthermore, it operates inside an OS-level sandbox on macOS, Linux, and Windows with granular workspace write, read-only, and network permission controls. The project is licensed under Apache 2.0 and is based on a fork of Codex's terminal surface. When bringing your own API keys or running local models, traffic stays private and communicates directly with providers without intermediate proxying.

Tags: ai agent, cli, coding agent, OpenSource, terminal

Visit Open Interpreter

Key features

  • Model-Specific Harnesses — Shapes agentic loops and requests for individual model architectures like DeepSeek, Kimi, and Qwen rather than using a one-size-fits-all wrapper. - OS-Level Sandboxing — Runs command execution and file edits inside a configurable sandbox on macOS, Linux, and Windows with workspace-write, read-only, and explicit network isolation. - Broad Model & Provider Support — Connects natively to OpenAI, Anthropic, Bedrock, Ollama, LM Studio, DeepSeek, Kimi, Qwen, and custom OpenAI/Anthropic-compatible endpoints. - MCP & Skills Integration — Inherits standard Model Context Protocol (MCP) servers, reusable skills, and AGENTS.md configuration files for repository-level guidance. - Session Management — Supports multi-tab workflows, allowing developers to resume (/resume), branch (/fork), or side-load (/side) active coding contexts directly from the TUI or shell. - Granular Approval Policies — Decouples sandboxing from human-in-the-loop approvals, prompting for verification before executing commands that escape defined workspace boundaries.

Use cases

Use Case: Sandboxed Test Fixing A developer can launch Open Interpreter in --sandbox workspace-write mode to diagnose and resolve failing test suites. The agent inspects local code and test logs, edits workspace files to fix the bugs, and prompts for user approval only when attempting network access or restricted shell actions.

Use Case: Local and Low-Cost Model Execution Engineers seeking to minimize inference costs or maintain strict data isolation can connect the CLI directly to local instances in Ollama or LM Studio, or to low-cost hosted endpoints like DeepSeek-V3 or Qwen3-Coder. Open Interpreter applies dedicated harness profiles to maintain tool-calling reliability with these models.

Use Case: Context-Aware Maintenance with MCP and AGENTS.md Teams with structured onboarding guides can define repository context in AGENTS.md and attach internal MCP servers for ticket trackers or databases. The agent loads these specifications automatically to execute recurring codebase maintenance tasks across separate terminal tabs using /fork and /resume.

Who it is for

Open Interpreter is aimed at software engineers, DevOps specialists, and security-conscious developers who want a fully scriptable, open-source terminal agent with granular OS sandboxing and flexible model choice. It is particularly well-suited for teams wanting to run cost-effective open-weight models or local inference without vendor lock-in. It is less suitable for users looking purely for in-editor IDE plugins or GUI-only code completion.

Install guide

Installation Install Open Interpreter using the official install script: bash curl -fsSL https://www.openinterpreter.com/install | sh

First Run Launch the interactive terminal interface from your project root: bash interpreter Within the TUI, select your model provider or configure project files: text › /model › /init To run with explicit sandboxing restrictions on file writes: bash interpreter --sandbox workspace-write

Example Session Run an interactive command to inspect and fix code: bash $ interpreter --sandbox workspace-write > inspect ./src/auth and fix the failing test in ./tests/auth.test.ts

Part of CLI Coding Agents

Categories: AI Coding Agents, Code Assistants, Developer Tools

Related CLI agents

  • Gemini CLI — Gemini CLI is an open-source terminal coding agent maintained under the Google Gemini organization (google-gemini/gemini-cli). It provides a command-line and…
  • Crush — Crush is an open-source terminal coding agent created by Charm (Charmbracelet), the team behind popular Go terminal UI libraries like…
  • Hax — Hax is a minimalist, native CLI coding agent created by Oleksandr Chekhovskyi. Written in C and designed for POSIX-compliant environments,…
  • amux — amux is an open-source agent harness and session multiplexer developed by Mixpeek. Built in Rust, it is designed to manage,…
  • San — San is an open-source terminal coding agent developed by the genai-io organization on GitHub. Built natively in Go, San is…
  • Pi Boss — Pi Boss is an open-source multi-agent orchestration extension developed by skyfallsin for the pi terminal coding agent ecosystem. It provides…
  • AI Tools
  • Categories
  • Industries
  • CLI Coding Agents
  • MCP Servers
  • MCP Categories