Repolode
Предложить
hydro13

hydro13/tandem-browser

An Electron-based, local-first browser where AI agents operate inside your real logged-in browser session via MCP or HTTP API.

Открыть на GitHubСайт
Звёзды
623
Форки
91
Открытые issues
21
Последний коммит
вчера

Что это?

Tandem Browser is a desktop browser built on Electron that puts an AI agent inside the browser itself rather than in a sidebar. The agent shares the same tabs, cookies and logged-in sessions as the user, reads the accessibility tree and DOM, inspects the live network log, and can inject scripts into the page. When a step needs a human, such as a CAPTCHA or a risky action, control hands back to the user.

Agents connect over MCP (Claude Code, Claude Desktop, Cursor, Windsurf and other MCP clients) or a 300+ endpoint HTTP API, on the same machine or remotely over Tailscale. Multiple agents can work in one browser with tab locks and explicit handoffs. The project describes an eight-layer security perimeter between web content and the agent, including domain blocklists, prompt injection defense and outbound credential scanning.

The browser also works as a daily driver, with workspaces, bookmarks, history, a local password vault, Chrome extension support and device emulation. It is in developer preview: macOS Apple Silicon is the primary platform, Windows 11 x64 is supported, and Linux is best-effort.

Ключевые возможности

  • 01Shares tabs, cookies and logged-in sessions between the user and the AI agent.
  • 02Exposes 257 tools over MCP and a 300+ endpoint HTTP API.
  • 03Lets multiple agents work in one browser with tab locks and handoffs.
  • 04Connects remotely over Tailscale without exposing the browser to the public internet.
  • 05Provides accessibility-tree snapshots with element refs for structured page interaction.
  • 06Includes an eight-layer security model with prompt injection defense.

Сценарии использования

  • Automating web tasks in SaaS tools that have no public API, using your own logged-in session.
  • Injecting custom UI changes or userscripts into sites while browsing.
  • Inspecting which API calls a page makes and what they return, locally.
  • Running an agent alongside your own browsing with human approval for risky steps.

Темы

#browser#electron#ai-agent#mcp#browser-automation#local-first#human-in-the-loop#typescript#ai#chromium

Частые вопросы

Что такое tandem-browser?
An Electron-based, local-first browser where AI agents operate inside your real logged-in browser session via MCP or HTTP API. Tandem Browser is a desktop browser built on Electron that puts an AI agent inside the browser itself rather than in a sidebar. The agent shares the same tabs, cookies and logged-in sessions as the user, reads the accessibility tree and DOM, inspects the live network log, and can inject scripts into the page. When a step needs a human, such as a CAPTCHA or a risky action, control hands back to the user.
tandem-browser можно использовать бесплатно? Какая у него лицензия?
tandem-browser распространяется с открытым исходным кодом под лицензией MIT. Перед использованием в коммерческих продуктах проверьте условия лицензии.
tandem-browser всё ещё поддерживается?
Активно поддерживается. Последний коммит в tandem-browser — 18 окт. 2026 г..
Для чего можно использовать tandem-browser?
Automating web tasks in SaaS tools that have no public API, using your own logged-in session. Injecting custom UI changes or userscripts into sites while browsing. Inspecting which API calls a page makes and what they return, locally. Running an agent alongside your own browsing with human approval for risky steps.
Где скачать tandem-browser?
Исходный код и инструкции по установке — на GitHub: https://github.com/hydro13/tandem-browser

Описание создано ИИ на основе README проекта. Статистика GitHub обновляется ежедневно.

Похожие проекты

apache / maka

Apache Maka (Incubating) — рабочее пространство для агентов с интерфейсами Desktop, TUI и CLI, которое записывает каждое сообщение модели, вызов инструмента и решение о разрешении в журнал событий только для добавления.

ИИ и машинное обучение#agent-runtime#ai-agent#llm
5,9 тыс.559TypeScriptApache-2.0Обновлён 4 часа назад