ComputerX draws praise for cutting busywork and enabling smooth, repeatable workflows—early users say it feels like a capable digital assistant that streamlines research, automation, and everyday tasks. Beta testers highlight rapid improvements, an intuitive web app, and a responsive team. Enthusiasm centers on chaining multi-step actions and tackling repetitive tasks, with several users citing strong results in research and marketing workflows. Critiques point to early friction, including confusing first impressions and hitting usage limits quickly. Overall sentiment is optimistic, with momentum and polish improving release by release.
Its cool to see how this tool helps you get things done on your computer just by typing what you need, which is very much needed in this fast moving tech world!
ComputerX
@khyatigupta Thanks! Turning plain English into finished work streams is exactly our play, zero friction, maximum velocity. Let us know what bottlenecks we should bulldoze next. 🚀
Quash
i love the tool, it gave me pretty good results for heavy research related tasks, pretty impressive work!
ComputerX
@donna_dominic Thank you!
How this is different from what Manus ai is providing ??
@omkaar_mahapatroI'm also curious about the same question!
ComputerX
@omkaar_mahapatro Thanks for asking — great question! While Manus AI focuses on general-purpose tasks, ComputerX is built specifically for automation and time-saving workflows. For example, you can build and schedule automations to track product prices or monitor company updates, which is especially useful for tedious or research-heavy use cases.
We're also expanding into deeper computer-level automation to handle tasks that current tools like Zapier or n8n can't easily manage. You can check out a quick demo of this in action here:
— it gives a glimpse into how our computer-use agent operates directly in a virtual computer to autonomously perform complex, multi-step tasks. This is the future of human–agent collaboration we're envisioning.ComputerX
@charlie_bakos Absolutely, computerx.ai is live right now. Use code LAUNCH30 for starter credits, and start shipping tasks in minutes. Your feedback will feed straight into the product roadmap, so fire away! 🚀
I may have rushed my reading is this full desktop automation or purely web?
ComputerX
@masikh Great question! This launched version is web-only. Earlier this year, we built a full desktop automation prototype — here’s a video demo:
. That said, the current computer use foundation model doesn’t yet meet the reliability and performance standards we’d need for a production desktop agent, so we’re actively researching how to improve its performance to make it truly usable.@sherryruan thanks for responding. It certainly feels like a more "difficult" problem to solve. The standardisation of APIs just isn't there. I did a rather large RPA project on a niche legacy windows app recently and I was just wishing there was an AI to sufficiently handle it the whole time. Had to resort to OCR etc at certain points.
ComputerX
@masikh Yup, API standardization is still lacking, and so much of automation today relies on fragile workarounds.
Today, the foundation model for computer use isn't fully matured yet — people are still working through challenges like accuracy, latency, and cost. But we're optimistic. With the rapid progress in VLMs, we believe the automation problem can be fundamentally solved by letting models operate on the front end like a human does — seeing, understanding, and acting directly on the screen. Exciting times ahead!
Hey Sherry, huge congrats on launching ComputerX! This is an incredibly ambitious and exciting vision for the future of work. The idea of a single, natural language interface to orchestrate complex tasks is the holy grail for productivity.
I was particularly struck by your commitment to transparency—the "No black boxes" approach. This is absolutely critical for building user trust, especially when the agent is performing complex actions. It's a feature many others overlook.
My question for you is about the delicate balance between simplicity ("Just ask") and control. As tasks become more complex, how do you plan to handle ambiguity in a user's natural language request?
For example, if I ask it to "research the best marketing strategies for a SaaS product," how does the agent decide what "best" means, or which sources to trust? Do you envision a more interactive, guided dialogue for these kinds of complex, nuanced requests, where the user can refine the agent's plan before execution?
This is a fascinating challenge to solve. You're tackling a truly foundational problem. Wishing you all the best with the launch!
ComputerX
@felix_foster That's a great question, and it's definitely something we think about every day as we see increasingly complex tasks coming in! The approach we're taking to avoid "black boxes" and promote user trust and control is by making the agent's thinking and actions visible at the UX level. Here are the key features we've implemented:
- enhance prompt feature - users can adjust enhanced prompts to clarify ambiguity upfront before sending requests to the agent
- ask questions - our agent proactively asks clarifying questions when requests aren't clear
- make the plan visible - agent provides a high-level plan so users can course-correct if needed
- making thinking & execution visible - agent verbalizes and shows all actions taken so users know exactly what's happening and can intervene if needed
Regarding your idea of "a more interactive, guided dialogue for complex, nuanced requests where users can refine the agent's plan before execution" - we actually implemented this in earlier prototypes involving computer-use technology (where agents use vision models to control a computer). For these tasks, we needed more precision since there's less room for ambiguity than in non-computer tasks. Here's a screenshot of what it looks like:
Back to your question about "simplicity" and "control." Currently, our agent primarily helps users with research and creating deliverables. Since these tasks aren't too high-stakes yet, we're emphasizing simplicity for a better user experience. However, if we see users requesting more complex or high-stakes tasks, or when we launch features with computer-use capabilities that require more precise planning, we'll consider adding more control features.
Like you said, it's a fascinating challenge to solve. Would love to hear more insights from you if you have any. Thanks!
@joyce_jiayu_he Wow, Joyce, thank you for such a thoughtful and transparent response. This is one of the best replies I've seen on Product Hunt.
Your four-pronged approach to balancing user control with simplicity (enhancing prompts, asking questions, and making the plan/execution visible) makes complete sense.
And thank you especially for sharing that screenshot! It's fascinating to see that you've already prototyped that exact "guided dialogue" for more precise tasks. It confirms that you're thinking several steps ahead about the user experience.
Your product philosophy of deliberately choosing simplicity for current use-cases, while being ready to introduce more control for future high-stakes features, is incredibly mature and insightful.
You've answered my question perfectly. I have no more insights to add right now, just a lot of respect for the way you're building ComputerX. Wishing you and the team all the best!
Really impressive speed on some of the tasks you showcase on your website!
ComputerX
@paul_hoeppner Appreciate the shout-out! Speed’s our north star.⚡