Tandem Browser is an open-source, local-first browser for AI agents. It is built for situations where an agent needs to operate inside a real browser session rather than outside it, sharing the same tabs, cookies, and logged-in pages as the person using the browser. The site describes it as a programmable workspace where human intent and AI capability meet.
Its capabilities center on letting an agent work with the live browser state instead of relying on a sidebar, screenshots, or per-site automation. Tandem Browser can read the accessibility tree, watch the network, work with the DOM, and access DevTools and live console output. It can also inject scripts to rewrite live interfaces, and it is described as able to hand back to the user when human input is needed. The page also says that any SaaS or web application that can be used in a browser can be used in Tandem Browser, without a per-site API or wrapper.
The browser is model-agnostic. It supports agents that speak MCP or HTTP, and the page names Claude, GPT, Gemini, OpenClaw, local Ollama, LM Studio, and custom scripts as examples. It can be used fully offline with a local model. The product also emphasizes co-browsing: the user and the agent are in the same browser at the same time, and the user can step in mid-task by typing, clicking, or switching tabs.
Tandem Browser is distributed as open source under the MIT license. The site also labels it a developer preview and gives its version as v1.11.0.
Tandem Browser sits in PulseGate's Browser DevTools & extensions category. It focuses on enabling AI agents to interact directly with live browser sessions for automation and research without RPA or per-site APIs. It is built as an open-source project for AI researchers, agent developers, automation engineers. Tandem Browser is open source under the MIT license. It ships for the web, macOS, Windows, and Linux, and it can be self-hosted.
Tandem Browser first shipped in 2026. Development happens publicly on GitHub with 559 stars and 569 commits in the last 90 days. Key capabilities include Local-first AI integration, co-browsing with agents, and open-source. It exposes integrations via an MCP server.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do