Solid - Agents with their own computers, accounts, and budgets.
Solid’s AI agents build apps, automate workflows and tackle work you lack the time or expertise for. They pick and set up their own real machines, create accounts and pay for services. They use your apps through APIs or log in like a person. They build what’s missing and fix problems along the way. Think beyond personal agents like Grok Bot and Muse: Solid is built for complex, long-running work for you and your team. Close your laptop. Solid owns the job from start to finish.
Replies
Solid
Hey everyone. I’m Trevor, founder of Solid.
We’re a team from MIT, Berkeley, and Waterloo, and we’ve raised $6M to build Solid around one goal: agents with human-level autonomy.
Many agents on the market can complete simple tasks. But when the next step requires resources they don’t have, like a server, an account, or a way to pay, they stop and wait for a human. They know exactly what to do next. They just don’t have the means to do it.
That’s the problem we’re solving with Solid: giving agents the resources they need and making sure they can use them securely and reliably.
For us, human-level autonomy means you can delegate a whole job to an agent as you would to a capable person. It figures out the steps, gets the resources it needs, and uses tools to deliver the result.
Our beta users have already used Solid to:
Build and deploy public-facing SaaS platforms for businesses across real estate, fintech, supply chain, and dozens of other verticals.
Automate raw material procurement by calling hundreds of suppliers and maintaining a live inventory dashboard.
Run a LinkedIn recruiting pipeline including search, outreach, and follow-ups.
We are building Solid to be the engine for actions, the way ChatGPT is an engine for answers. Anyone can use Solid to get work done, or wrap that engine in their own products to create customized agents.
Intelligence is becoming a commodity. The power to act on it should be too.
@tkeith looks great, but how readable is the generated code when if I need to hand it off to a dev team?
Solid
@priya_kushwaha1 Good question - in general the code is surprisingly clean! It's important not only for hand-off to dev teams, but also to keep the generated products maintainable as they grow. Many of our users have created 300k+ line codebases, which is only possible with cleanly architected code.
@tkeith Perfect thanks, makes it much easier to trust it for long-term projects. best of luck for the launch
Pandada AI
Glad the support example includes checking back with the customer. I'd want to know the fix actually worked for them before calling it done.
Solid
@panwangqun Yes, you can always teach Solid what works best for your business!
Can Solid help write the test scenarios for an AI product, or do I need to prepare those before it can run evaluations?
Solid
@eren_mindful Yes! We have users using Solid to create evals for their AI data query product - so Solid would log into the browser to use their product and query 500 questions then compare the answers to decide if the product quality slips after each update. This process would run for over 10 hours. This is a great example of how Solid can handle complex, long-running tasks.
Ada.im
What would be a good operations task to start with? I'd like something with a result I can check.
Solid
@s_cen Build a dashboard is great. Give it the information you want included and the access it needs, then check the result and ask for changes. You can try the "Agent as a service" option - instead of having static code under the web interface, you can have an agent go and execute a task and update your dashboard numbers. Solid is great for that.
Finally someone tackling the last mile problem. Building is easy now. Maintaining and wiring everything together is the hard part.
Solid
@henryevans11 Exactly. It turns out the last-mile problem is not an intelligence problem. It's whether the agents have access to enough resources (computers, accounts, budgets). After that, humans are no longer blocking agents, and we can finally maximize their productivity power.
The concept of this tool sounds very interesting just want to know that for how long can a solid run on a complex task without needing human intervention.
Solid
@jitendra_bhatt let's see... And now we have a lot of users doing mechanical part assembly design on Solid. Those jobs can run for over 6 hours. We also have users running their evals on Solid, which can take over 10 hours.
Solid is extremely good at handling complex and long running task without human intervention because of all the tools it has access to.
1752vc Pitch Deck Analyzer
This is great. Full operating system in my hands.
Solid
@tmaleh_ Exactly!
Great job Solid team :). Love the video.
Seems like it is as simple as Lovable, but as powerful as Replit in terms of infrastructure. Does Soild sandboxes have access to Mac boxes too? As in can you build a iOS mobile app? Or a desktop app?
Long term, how do you see it being different than Lovable? Seems like you need to get the solid subscription and can't use your existing claude/codex/grok/etc subscriptions?
Solid
@michael_yagudaev Thanks Michael! I would say we are more powerful than replit, sometimes even Claude and Codex (feedback from our users). We are already supporting Mac mini and iphone, but it's not public yet, you can contact the team if you need one!
Right now we have users using their claude/codex subscription in the cloud machines, you can also use any of the above agents by downloading them to the cloud computer or through browser. We are also planning on allowing user to just use their existing subscriptions too, if many users ask for this.
LowTech AI
This is epic!! The last few things I have to do by hand, no more.
So if I understand correctly the agent can literally spin up a VM and create a SaaS account if needed? That is wild. How do you manage security permissions for that?
Solid
@albertnelson that's correct. Agents can sign up just like a normal human, and you manage security permission just like you would with a normal human. You can set special access and permission for the agents, and you can tell them about the policy and rules. With the latest LM intelligence, they are very good at following the rules.