Ready for the Next Wave

The world of AI agents is changing quickly. And I can count on one hand the number of days I have not woken up, seen something new, and asked myself, is MCProspero still relevant? I think most founders feel this way, generally, but this space is on a tear…

And the answer is yes. Why? A few things.

Most importantly – and I experience this every day – LLMs are unpredictable. I don’t just mean they hallucinate – sometimes they do – but they often try to be too helpful, or try to achieve a goal in a creative way, that is not what you want. And when it comes to securing your data and your reputation, this is a big problem. You need control points (guardrails) that work within the reasoning steps of the agent – “I am about to make this decision, is that ok?”

This needs to be architectural, intentional, and not an afterthought. And as you can imagine, the LLM providers are not particularly incented for their platforms to do this.

So if that’s important to you, as it is to me, MCProspero is still very relevant.

Second, we’re not tied to any stack. MCProspero is built on MCP – an open standard. Use Claude to create your agents, run them on GPT-4, monitor them from ChatGPT. Or use all Anthropic. Or bring your own model. Again, the LLM platform providers are not particularly incented to make this easy. I think it needs to be easy.

Lastly, this is for everyone. Not just developers. Simple agents helping humans get control of their lives back.

So anyway, go crazy with Claude Routines, or OpenClaw, if that’s your jam. But me, I am going to keep trying to make MCProspero useful for normal people in a safe easy to use way.

That said, I need more people using it to really vet the idea further. So I am especially excited that we can let the next wave of users in.

Here’s what we’ve been building to get ready.

MCProspero has always been conversation-first. Describe what you want an agent to do, Claude helps you build it, MCProspero keeps it running. That hasn’t changed and it won’t. But I’ve known for a while that we needed something visual too.

Some things are just better as a dashboard. “Is everything ok?” shouldn’t require opening a conversation. You want to glance at your phone, see green lights, and move on. Or see a red light and know exactly which agent needs attention. So we built a web dashboard – fleet view, agent controls, settings, the works. Start, stop, approve, run – all right from the browser. It’s not replacing the conversation. It’s the companion to it.

The auth story needed to get a lot better too. We started with Google and GitHub sign-in, which works for a lot of people. But not everyone. Some folks don’t have Google. Some don’t want to use their GitHub login for something that isn’t a dev tool. We needed email-based login – just type your email, get a link, click it, you’re in. Magic links. What seemed like a small addition turned into a bigger project (a full auth stack evolution), but the end result is cleaner and simpler than what we had before. Three ways to sign in, zero friction.

And I also needed more tools for keeping track of more users, system health, etc. Not glamorous work, but the kind of thing that makes the difference between “hobby project” and “platform real users can depend on.”

So come brave souls! Help make this platform better for everyone. If you’re keen to give it a try, sign up here: https://tally.so/r/QKrrOA

-greg