Today we’re launching Simular Pro - our first production-grade computer use agent that can use computers like a human.
We built Simular Pro because most AI tools stop at text or APIs. In the real world, workflows often live inside GUIs - spreadsheets, EMRs, CRMs, dashboards - and they change all the time.
Simular Pro currently runs on your mac or macOS virtual machine. It can see the screen, move the mouse, click buttons, type text, and adapt when the UI changes.
How we built it:
Most agents rely purely on LLMs - great explorers, but flaky executors.
RPA tools are stable, but rigid.
Simular Pro combines both to solve the Generality vs. Repeatability problem:
• Neural Agent: explores and adapts.
• Symbolic Agent: locks in what works and executes.
Every action becomes editable, deterministic code — fully transparent and repeatable.
Examples that Simular Pro can do:
1️⃣ Price comparisons on Labubus for my daughter’s birthday → 30 legit options ranked by value in a spreadsheet.
2️⃣ Finding YouTube influencers for Simular’s launch → keyword search, pulled subs/views, ranked top 60 in minutes.
3️⃣ Batch drafting cold emails → pulled leads from CRM, grabbed context, drafted tailored intros; I just reviewed & sent.
4️⃣ Social content from an 18-page paper → drafted an X thread, suggested visuals, loaded into NotebookLLM to spin up a podcast.
5️⃣ Hiring an ex-DeepMind PM → ran LinkedIn search, filtered results, pulled names/roles/schools into a spreadsheet.
6️⃣ End-to-end Xcode build → compiled & packaged app, zipped & uploaded to Google Cloud with auto-update logic.
We’ve seen business customers automate everything from insurance claims processing to healthcare admin to data migration - all without custom integrations.
We’d love your feedback, ideas, and wildest use cases. Drop a comment if there’s something you wish an AI could just handle for you.
Realistic simulations powered by advanced AI. Train, test, and explore scenarios with precision and ease. Bringing virtual experiences closer to reality for smarter decisions.
Actually pretty handy for Mac folks who hate repeating the same clicks and typing. Just tell it what you need in plain English, and it mimics the steps-set up a routine task in,like 10 minutes.
This sounds incredibly powerful, especially the approach to combine neural and symbolic agents to tackle the generality vs. repeatability challenge! That's a huge pain point for automation. I'm curious, how does the symbolic agent "lock in what works" to ensure consistent execution even with minor UI variations or updates? Also, any plans for a Windows version in the future? 🤔💻 @Simular
This looks incredibly powerful! The hybrid Neural + Symbolic agent approach is a brilliant solution for the repeatability problem. I'm curious about the 'editable code' part. How much technical skill is needed to troubleshoot or customize a workflow if the agent gets it wrong? Is it still user-friendly for non-coders? Congrats on the launch! 🤖⚙️
Brilliant! Combining the strengths of LLM-based agents and RPA tools while mitigating their weaknesses unlocks massive value for users. Can't wait to start using this!
It seems like a great idea. I wanted to install and try it, but sadly, I don't use a Mac.
Really impressed by how Simular Pro bridges the gap between LLM flexibility and RPA stability. The combination of neural + symbolic agents feels like a breakthrough for automating complex, ever-changing GUI workflows.
Congrats on the Launch!!
Simular is pioneering a bold shift: autonomous computers that use software like humans do. From spreadsheets to creative tools, it’s not just automation, it’s software that operates itself. A glimpse into the future of human-computer collaboration.
This is truly fascinating, @Simular ! The 'Neural Agent + Symbolic Agent' approach to solve the Generality vs. Repeatability problem is a brilliant architectural choice. It makes total sense for dynamic GUI environments. The Xcode build automation example really highlights the depth of what Simular Pro can achieve. My main question relates to security: Given that it interacts directly with local applications and potentially sensitive data, what measures are in place for data privacy and access control? 🤔🔒
this is actually fire - the neural + symbolic approach is exactly what's been missing in computer use agents. most tools are either too exploratory (pure LLM) or too rigid (traditional RPA).
curious about the macOS limitation though - any plans for linux support? also wondering how you handle the deterministic code generation when the UI changes significantly vs minor layout shifts
