Skyfire: Building out payment rails for AI Agents

Skyfire: Building out payment rails for AI Agents

I wanted to make sure I documented skyfire.xyz here on Conversational AI News. The company's mission is one that is absolutely fascinating. And here is the Overview from their about page:

To create the world’s first fully autonomous, open economy for Agentic AI. AI Agents shouldn’t need banks, credit cards, or even humans to get stuff done. To accomplish this mission AI Agents must be:
1) Endowed with an identity that is trusted, verified, and unique , 
2) Empowered with financial rails that allow them to make and receive payments globally and instantly
3) Governed by business rules, logic and budgets, defined by the enterprise and customers,
4) Able to discover and hire resources to accomplish increasingly complex tasks without the need for human intervention or negotiation.  

Now, you might argue that one option or one way of going about AI agent payment management is simply to have them use your own credentials (e.g. "I've stored my card number in the .env file, please use those details.") However, I think that this is to misunderstand the possibilities and potential of the new awesome economy that may well present itself for us all.

On the face of it, it's actually quite simple at the moment, you just give the relevant tokens to your various different agents, scripts or automated systems and they go ahead and use your systems and your access. very quickly, though that can get quite annoying and rather inefficient.

And the last thing you want to be doing is having your agent burning CPU time, memory, credit, trying to figure out how to log in with your credentials and charge your credit card for access to something that needs to get done.

So for example, right now I have a little agent that helps me understand what route I should be taking to the children's school every morning. (Let's just pretend it's an actual agent and not simply a little script that's running on a cron job. It sounds better!) that agent, script is using one of my Google Cloud Console API keys for access to query Google Maps.

I had to do quite a lot of messing around, cutting and pasting in order to get that little API key for this particular reality. I need that to be seamless, and we're going to want that ongoing to be totally seamless. So I see where the Skyfire team are coming from.

Let's have a look at a few more screenshots.

So this one's interesting as well because what they're highlighting here is my agent's scripts or automated whatever's are going to ideally want to be able to consume all sorts of services easily and effectively and seamlessly. Now, if you can make that work really, really easily, then my agent could well (under my direction or otherwise) start consuming those resources.

I could imagine, for example, authorising my agent to use Skyfire and then asking it a whole lot of questions like, for example, go and book a flight – and then it needs to go and actually do that. If there's a beautifully formatted by-flight API offered via Skyfire, then that would make a lot of sense for me to be encouraging or requiring my Agent(s) to use this.

And then when it comes to identity, I think that's another really important thing to begin to wrap your mind around. I think the default way in the ultra-short-term is simply it's my agent, my responsibility, it's doing what I specifically asked it to do, even if that's not the case, it's still fundamentally my responsibility.

But as things progress, we're going to want to have agents with individual identity, simply, as the team point out in the above screenshot, for the purposes of authentication and authorisation.

I think there's also going to be a lot of value in verifying your identity or the agent's identity and by abstraction, your identity. Being first and being efficient could very well really help the team at Skyfire to drive things forward brilliantly.

This is certainly one company to watch.

Good luck team Skyfire.

Read more