Introducing Simular

A browser agent for real-world tasks

We all spend too much of our lives in front of computers—whether for work or everyday tasks. Imagine a future where we’re no longer glued to the keyboard and mouse. That’s the idea behind Simular, which we founded in 2023. Today, we’re excited to launch Simular, our first browser agent for macOS, a step closer to making that dream real.

Back in the late 2010s, my co-founder Jiachen and I were working on multi-agent systems at DeepMind. We realized that general AI won’t emerge in a closed lab—it must benefit people in the real world to improve over time through real interactions. That’s why we built Simular to handle mundane computer tasks so you can spend more time with family, out in nature, or doing what you love.

Our work’s showing real promise—our open-source framework, S2, outperformed Anthropic and OpenAI on real desktop tasks (OSWorld benchmark) and smartphone interfaces (AndroidWorld). From there, we created the Simular browser agent—a fast, dependable agent that teams up with humans to get things done. It’s already set records, achieving 90.05% on WebVoyager*, surpassing other browser-based systems, including OpenAI’s Operator.

Native, fast and dependable

Most agents today run in cloud-based virtual machines. While powerful, this setup is expensive, slow, and detached. It’s difficult for users to interact with the agent’s actions. Logging into your accounts becomes cumbersome. And ultimately, the whole experience feels disconnected.

We designed Simular to be a desktop agent with an embedded native WebKit browser rather than an agent running on the cloud. This means enhanced security, faster performance, and a seamless, collaborative experience between humans and the agent.

‍Fully autonomous agents aren’t yet human-level—not consistently, not reliably. Pretending otherwise only leads to frustration. But for many repetitive, mundane, or unfamiliar tasks that we would otherwise not do ourselves, they can already produce usable results.

That’s why we built Simular to do something more grounded: provide a fast, secure, and local environment where agents can act autonomously—while users can step in at any time. 

It’s designed for seamless human-AI cooperation, not replacement. Real users need more than raw capability—they also need transparency, responsiveness, and control.

Human-AI collaboration

Simular defines a new kind of human-AI relationship:

• On-device autonomy: The agent runs in a local, sandboxed browser. You don’t need to trust the cloud with your credentials, and your sessions stay entirely on your Mac.
  

• Shared control: You can use the browser while the agent is working. Click links, edit forms, or jump in mid-task—no handoffs needed.
  

• Secure and familiar: Since the browser is native, it behaves just like Safari. You can log in, open tabs, and use all your usual workflows—with an agent quietly helping behind the scenes.
  

Simular in real life

From simple errands to more complex tasks, Simular is designed to work alongside you, enhancing how you use your computer without getting in the way. Below are a few real-world examples to show what’s possible. These aren’t just demos—they’re meant to inspire how Simular can fit into your own workflow and daily life.

Need help with party planning? See how Simular helped create an Easter shopping list, add all the items to an Amazon cart, and ask for human permission before checking out.

Tariffs are making car buying tricky. Watch Simular give actionable advice on buying Japanese and Korean cars, pulling from real-time tariff updates.

Scanning news can be time-consuming. Simular grabs trending events from news sites and creates a podcast via NotebookLM in minutes. See it in action!

Hate clunky government portals? Simular navigates them smoothly and checks your tax payment status with ease.

These are just a few snapshots of what’s possible today. As Simular continues to evolve, so will the range of tasks it can take on—making everyday computing faster, smarter, and more collaborative. We’re just getting started, and we can’t wait to see what you’ll do with it.

Roadmap

We designed our first product as a browser agent because that’s where most of people’s computer work happens today. But this is just the start. Looking ahead, we’re working to bring Simular to other systems like desktops and smartphones.

For a deeper dive into the technology behind our computer-use agents, check out our technical blog post: https://www.simular.ai/articles/agent-s2

‍Building autonomous computers doesn’t mean replacing humans. It means cooperation.

Free your hands from the computer. Download Simular today for free. 

*Note: The webvoyager result is achieved by the multi-agent mode.