Agentic AI — AI that can not only write code but execute it, observe the results, and iterate — is the most significant development in developer tools since GitHub Copilot. The difference between an AI that writes code and an AI that runs code is the difference between a consultant who gives you advice and a contractor who delivers a working product. LumiChats Agent Mode is in the second category: a full Node.js environment running in your browser where AI code runs in real-time.
How Agent Mode Works
LumiChats Agent Mode uses WebContainers — a technology that runs a full Node.js runtime in the browser via WebAssembly. When you activate Agent Mode and give the AI a task, the following happens: the AI writes code based on your request, the WebContainer executes that code in a sandboxed environment, the AI reads the output (or error), adjusts the code, and runs it again. This loop continues until the task is complete. All of this happens in your browser — no server costs, no security risks from running user code on a remote machine.
What You Can Build in Agent Mode
- Data processing scripts — upload a CSV or Excel file and ask the agent to clean it, analyse it, and generate a report. The agent writes the Node.js script, runs it on your file, and delivers the output.
- Web scraping — provide a URL and ask the agent to extract structured data. The agent writes a Cheerio or Playwright script, executes it, and returns the results.
- File generation — ask the agent to generate a Word document, Excel spreadsheet, PDF, or PowerPoint presentation. It writes the code using appropriate libraries, generates the file, and provides a download link.
- Full-stack mini-applications — describe a small web app and the agent will build the HTML, CSS, and JavaScript, execute it in the WebContainer, and return a working preview.
- Code refactoring from your codebase — upload a ZIP file of your project and ask the agent to refactor, add a feature, or debug a specific function. The agent reads your code, makes changes, and runs the tests.
Agent Mode for B.Tech and Computer Science Students
For engineering students, Agent Mode is the most useful tool for project work and assignments. Final year project data analysis, database queries, algorithm implementation, and API integration tasks — all of these involve write-run-debug cycles that Agent Mode handles autonomously. The practical experience of watching an AI debug its own code in real time is also one of the best ways to learn software engineering — you can observe the debugging process, ask the agent to explain each step, and understand why certain approaches fail before being corrected.
Pro Tip: Agent Mode requires a paid day pass (it is not available on free tiers). For developers, the ₹69/day cost is typically recovered in the time saved on a single complex debugging or data processing session.