Get started in seconds.
Download the IDE or install the CLI — both bring full AI power to your workflow.
Download the IDE or install the CLI — both bring full AI power to your workflow.
Use any model in Plan or Execution mode — even open source models that don't natively support tool calling.
Plan mode lets you map out your entire feature before writing a single line — define tasks, set context, and hand off to the agent with full confidence.
Execution mode is a focused version of Agent mode — it follows your plan step by step, writing code, running tests, and committing changes autonomously.
Open source. Plan-and-Act. Every model. No vendor lock-in.
Edit code inline using natural language. Select any block of code, press a shortcut, and watch the AI modify it in real-time.
Review all AI-generated changes side-by-side. Approve, reject, or iterate on edits line-by-line before committing them.
Experience fast, context-aware code suggestions as you type. Predicts your next lines of code with ultra-low latency.
Map out complex features step-by-step in Plan mode, then hand over to Execution mode to run tasks autonomously.
Let the AI run terminal commands, execute test suites, analyze compiler output, and self-correct errors as they arise.
Bring your own API keys. Toggle between Claude, GPT, Gemini, DeepSeek, or run open-source models fully locally.
Ship `.loophole` or `.clinerules` with your repo to teach the editor your project structure, conventions, and styling guides.
Search and index your entire codebase dynamically. Plug in Model Context Protocol (MCP) servers to read databases and APIs.
Loophole works with every major provider. Bring your own key, your own endpoint, or your own weights and keep your workflow exactly the same.
