Errand — AI browser agent

Errand is an AI agent that does web tasks for you, right in your browser. Tell it what you need in the side panel: "fill in this form from my notes", "compare these two products", "copy the table from this page into a spreadsheet". Errand reads the page, moves a visible cursor, clicks and types, and reports back. You watch every step and can stop it at any time. BRING YOUR OWN MODEL Errand works with any OpenAI-compatible API. Use OpenAI or OpenRouter, or keep everything on your own machine with Ollama or LM Studio. Built-in presets fill in the details; paste a key, test the connection and pick a model. There is no Errand account and no subscription. YOU STAY IN CONTROL • Errand works only in its own purple "Errand" tab group. • It asks before working on a site for the first time (allow once, always, or deny). • It asks before risky actions: purchases, submissions, deletions, password fields and downloads. You choose which words count as risky. • A live cursor and highlights show exactly what it is about to do. • Stop ends the task immediately. PRIVATE BY DESIGN Errand has no servers, analytics or telemetry. Settings, keys and chat history stay on your device. Page content is sent only to the model provider you choose, and with a local model it never leaves your computer. FEATURES • Attach images, PDFs and text files to a task • Works across multiple tabs • Chat history stored locally • Light and dark themes • Free and open source (AGPL-3.0): https://github.com/Dread63/errand Note: while a task runs, Chrome shows a "started debugging this browser" bar. That is how Errand sends real clicks and keystrokes; closing the bar pauses the task.