DOZ vs Tandem Browser

DOZ is a Chrome extension that follows natural-language instructions for browsing websites and handling email. It can navigate pages, click controls, scroll, type text, press keys, search, and fill forms. Through Gmail and Outlook/Hotmail APIs, it can read, search, reply to, and compose messages. DOZ says it understands English and Arabic. Users can choose full automation, confirmation before actions, or plan-only mode. A step timeline displays screenshots captured before and after actions. Supported AI providers are DeepSeek and NVIDIA Nemotron, and the extension requires your own API key for one. The Free plan costs nothing, with third-party API charges your responsibility. DOZ says page content is processed in the browser and sent directly to the selected provider, while keys, settings, and email OAuth tokens are stored in Chrome storage. Installation uses a GitHub download and unpacked Chrome extension loading.

Tandem Browser is a free, MIT-licensed browser in developer preview for running AI agents alongside a person’s browsing session. Agents share tabs, cookies, and the DOM, and can work with the accessibility tree, rendered DOM, network requests, console output, and DevTools. If an agent reaches a CAPTCHA, unclear form, or judgment call, it pauses for user help. MCP- or HTTP-speaking agents can connect locally or through a private Tailscale network, with separate workspaces and tab locks. The documented API includes more than 280 HTTP endpoints and 257 MCP tools, protected by bearer tokens. Tandem has no cloud backend, subscription, or telemetry according to its FAQ; browsing data remains on the machine unless sent elsewhere. It can run offline with Ollama or LM Studio. Daily-driver features include extensions, bookmarks, history, a local password vault, workspaces, and sidebar apps. It uses Chromium and supports macOS Apple Silicon, Windows 11 x64, and Linux on a best-effort basis.