Asteroid lets anyone easily build complex computer-use AI agents. Both non-technical users and engineers can now reliably automate browser and computer tasks 20x faster with Asteroid’s AI agent builder. It’s used in production across hundreds of startups and enterprises, primarily to solve healthcare interoperability challenges.
This is the 2nd launch from Asteroid. View more
Asteroid
Launching today
Asteroid lets ops teams and engineers build computer-use agents for browser, Linux, and Windows workflows in minutes. Our meta-agent, Astro, builds the agents, writes scripts as it goes, and makes repeat runs faster and cheaper. Last month, Asteroid agents completed 150,000+ executions across EHRs, benefits portals, insurance carriers, Citrix, desktop apps, and VPN-protected environments.
Free Options
Launch Team / Built With
Asteroid
👋🏻 Hi Product Hunt! I'm David, co-founder of Asteroid.
We spent the last year perfecting the experience of building computer use agents.
RPA, Playwright, and browser automation are as old as bad software itself. The problem got bigger with AI agents: they're useless if they can't connect to systems that lack APIs. But every computer use agent framework until now has overpromised and underdelivered.
Thankfully, 2026 LLMs like Claude Opus 4.6+ make this hard problem actually solvable.
We went deep on complex customer use cases and took a clear stance: it's better to automate 50 complex browser portals extremely reliably than 1,000 at surface level. We focus on repeatable workflows, not one-off tasks.
A year of iterations, major updates, and small shipped features later, we have a full-stack platform that combines everything you need to build, run, and optimize computer use agents.
Just last month, Asteroid agents completed 150,000+ executions across 1000s of EHRs, benefits portals, insurance carriers, Citrix, desktop apps, VPN-protected environments, including:
We're the only platform supporting:
🌐 Any environment: browser, Linux, and Windows desktop
🖱️ Any interaction mode: visual computer use, structured DOM, or scripted execution
🧠 Any frontier LLM: OpenAI, Anthropic, Google, and more
🤖 Meta-agent: builds and debugs your agents for you
💾 Persistent filesystem: each agent stores scripts, files, and memories for reuse across runs
🔐 Session persistence: agents stay signed in and pick up where they left off
📬 Email inboxes: every agent gets its own inbox for 2FA and notifications
🕵️ Proxy support: stealth browsers, residential proxies, custom VPNs, and Tailscale
👀 Live view & recordings: watch runs in real time, review sessions, and add human-in-the-loop checkpoints
🛡️ Secure & compliant: HIPAA and SOC II Type II certified
🧩 API-first: webhooks, MCPs, and SDKs out of the box
You need all of these together to deploy computer use agents across critical industries at scale. Our use cases inlcude:
🏥 EHR / Patient Record Extraction: hospice and post-acute patient ingestion across agency EHRs
🩺 Voice-to-Form Clinical Intake: GP receptionist call transcripts turned into submitted triage forms
💳 Health Insurance Enrollment: ACA enrollment, plan handling, CRM sync, and binder payments
📅 Patient Creation & Scheduling: ECW patient creation, referral intake, approval checkpoints, and scheduling
📄 Insurance Policy Management: creating, filling, updating, and endorsing policy forms across carrier and agency portals
💬 Insurance Quoting: submitting quote requests, configuring coverage, and extracting results from carrier portals
🔁 Insurance Renewals & Endorsements: initiating renewals, applying mid-term changes, and tracking endorsement status
🧾 Insurance Claims Submission: end-to-end claim entry via Citrix VDI and status tracking across portals
🗓️ Intelligent Scraping: schedule extraction across 10,000+ venue calendars
📝 Form Filling: government portals, banking dashboards, and internal finance tools
If you've used Cursor, Lovable, or Claude Code and felt that magical moment, you'll feel it again when you see Asteroid automating portals you'd already given up on.
Try it now here or book a scoping call to talk through your use case.
Use promo code ASTRO10 for $10 in free credits to get started.
Would love your feedback! 🚀
Asteroid
@pie3 It's been so much fun working on the Asteroid platform for the last year or so! We've built something epic!
@pie3 love that you included live view and recordings with human checkpoints. for mission-critical workflows like patient scheduling or insurance quoting, you absolutely need that audit trail and the ability to intervene. very thoughtful build.
Asteroid
@priya_kushwaha1 you're 100% right, especially healthcare customers use the live view and HITL functionality a lot to confirm high risk actions like patient updates.
I've used several Agents, specifically browser agents, over the past year and found Asteroid to be the best one. It was actually able to visit sites that require cookies and managed some of the harder to do things I wanted. Excited to see that the team is now going for computer use agents!
Asteroid
@riccardo_varenna1 Thank you Riccardo! Computer Use agents definitely present different challenges to the browser stuff that we've been iterating on for over a year. Not having the DOM for example makes it a more difficult environment to work in. But we've seen great success so far on Windows and Linux tasks:)
Excited for this! Curious, how would Asteroid work for businesses using legacy systems that are gated (ie Iconote for Rehab Clinics)? Is there any vertical or situation where it can’t be used?
Really stoked to try this out. I’ve been selling agents to mid-market/enterprise businesses and the biggest bottleneck has been gated legacy apps.
Asteroid
@jaykhan3 We have many customers across healthcare that use our browser agents across legacy systems that are even behind VPNs, so I'm sure Iconote for Rehab Clinics would not be a problem. I'd not say there is a vertical Asteroid can't be used, one thing we don't do yet is on-premise deployment. I'd recommend building an agent and you will see - https://platform.asteroid.ai/
Asteroid
@veerhunt_agai you can configure that yourself - using prompts or deterministically. You can tell the agent "when failing, try again" or "when failing, ask user for help". We also have fancy logic around detecting selectors on the page which can trigger human intervention. This is super useful for critical applications!
Asteroid
To give you some stats from our small engineering team, over the last 12 months
we've shipped 3,093 pull requests
we've closed 1,351 Linear issues
we've scaled our infrastructure to support over 100x the volume
It's been a long journey, but I'm incredibly proud of what we've built and I'd love you to check it out!
Asteroid
@joe_hewett A* team of builders
Healthcare interoperability is exactly where computer-use agents shine - EHRs don't expose decent APIs. How does Astro handle sessions that need MFA or CAPTCHA mid-flow? That's typically where automation gets flaky.
Asteroid
I'm so proud of the work we've achieved as such a small team. I joined right after the last Product Hunt launch, and looking back I can't believe just how much the platform has changed. The way it looks, and what it offers. The things our agentic workflows are able to do for customers are truly game changing. So many people now rely on automations running on Asteroid, and they won't even know it! Unless they stop and think: "Wait, I got that GP referral so quickly this time around". Let's go! ❤️🔥
Asteroid
@davide_lct it's incredible how much a small team can ship and how many customers it can serve, when motivated and with high agency!