INTRODUCING: @doanythingapp Do Anything Agents are a totally new kind of agent They: - work independently for weeks/months+ - have their own email - self manage entire projects - can use almost any tool on the web Today the alpha opens to the public. Here's how they work:

Unlike other agents that need you to babysit them, Do Anything agents work completely autonomously to finish your project. Your prompt can be small: Get me a meeting with a Stripe engineer Or very big: Start and run a coffeeshop in SF

Do Anything agents operate 100% in the cloud. Our goal is that your Do Anything agent acts like the best employees in the world do. You can chat with them in our app & website. You can also call/text/email each other. They do a great job learning how you like to be contacted.

This is all possible thanks to a new long running agent framework called MEESEEKS. It is a collection of simple but important agent abstractions that allow an agent to work like a human to manage long running projects instead of just tasks.

One of the most important abstractions is the ability to self reflect on its work. This allows the agents to be critical of its own work & continuously learn best practices over time. It also gives the agent freedom to try things that might not work & keep trying if it fails.

Sleeping is another extremely important abstraction for long running tasks. Sometimes, you need things need to happen in the world before it makes sense for the agent to do anymore work. Sleeping allows our agents to keep monitoring a project without spending a ton of tokens.

But the most important feature of a Do Anything Agent is it's independence. Other agents take action AS you in YOUR accounts. Do Anything has its own identity. It takes action as ITSELF in ITS accounts. This is why your agent has its own name and email address.

This is extremely important. If you onboarded even the best human employee, you still would not give them access to your accounts. The reputational risk is too high. But you would give them lightly associated accounts. This allows agency without risk. Agents are the same way.

While they can use your tools using our 3000+ api connections, they prefer to use their own browser and email to open their own accounts where possible. This makes it easier for them to manage their own accounts and passwords, keeping you free from having to babysit them.

Our agent framework makes it super easy to add many tools and capabilities. In the next few weeks we will add bank accounts and expense cards (requiring your approval or a set budget to transact) I will also be adding agent side notifications, allowing the agent to set listeners into apps to get "woken up" by them.

We are still in alpha. I would compare this release of the Agent (0.6.2) to 2019 Tesla self driving, it will get you were you need to go, but every once in a while you might be needed to help. If you encounter something the agent can't do, submit a bug report and we will try to fix it in the weekly release (Sundays).

Right now, the Do Anything agent is like a VERY good & cost effective VA. But by end of 2026, my goal is that Do Anything agents are "one-shotting" profitable businesses at 5% efficiency, on par w/ humans. I want each user to have an army of CEOs in their pocket.