The Agent Meta - Introducing NEAR AI, OpenClaw, and IronClaw

Captions as YouTube publishes them, usually machine-generated, so product and people names can be misheard.

1. 0:00

   Uh yeah, so I'm going to kind of go over a couple different things here regarding AI and developers and introduce uh Open Claw as well and how that works with Near AI. Uh so first uh with the new agent meta here and working with AI as developers and developer platform at NearDev, we've had to make sure that we've actually address that as well.

2. 0:22

   That all fits in together with what we're doing here with agents. So from near docs and what we consider AI grounding is that we have a problem that we've been trying to solve since pretty much the you know fall of last year. Uh docs are for humans and they've been written that way always. And so LM struggle with the ambiguity in that document and sometimes when you're just trying to show examples you're not being

3. 0:45

   always uh emphatic and clear. So what happens is LM can't process that. So we need to actually better address docs in a way that allows uh devel LLMs to process that information to make better code. So we want to eliminate the frustrated developers and broken code.

4. 1:03

   And so what we have done is what's called GEO. So it is kind of like SEO but now this is actually for generated output. And what we do for that now in our documentation we've had this actually before and all these other chains kind of promoted stuff. We've had this actually since October where we've structured metadata for AI retrieval. Uh we have deterministic answers. If you

5. 1:25

   see now in our docs, you'll start to see more and more of like this is how something is like very clearly how things work. And then we've also started working on making sure through the community we have agent skills as well that you can actually just download and input into your agent and they can use that. So if you just point your agent to docs.ner.org you will have the right information for you to start actually building easily on

6. 1:49

   near with your agent uh from there and we'll continue to update those as we go. But what's really exciting is the whole agent thing. If you haven't seen this, I don't know where you are online, but this is everywhere um across the stack.

7. 2:02

   And open claw while it is open source and it has been uh really widely touted everywhere uh it is problematic for a few different ways and right now we know it's a black box and this is all stuff that near AI is addressing uh your prompts go to servers you can't verify your data process environments you don't control all that stuff that we've talked about hundred times about near AAI but with this of the private inference the verifiable execution

8. 2:32

   user ownership. It really supercharges Open Claw with allowing your agent to now be a private box for your information. Allows you to then have a situation where all the requests are being run inside that TE that trusted execution environment and then you know that where your information is going and staying. It is not foolproof and that is why you probably have seen previews of Ilia talking about ironclaw which we're

9. 2:59

   not going to talk about today. But this is a great step forward from what the traditional just setup of here's open cloud download and load it on your VPS or Mac mini. Um this is straightforward again all the hardware security is still there. It is just as if you were having the chat with Near AI is built in now to this actual agent yourself. So you're still having all that same trusted execution through all your requests to your AI.

10. 3:28

    And what we now have is in openclaw that agent gateway here is whether you're still using it just like you see online people talking about through telegram through discord signal whatever um and then you can still use your different models. Now if you don't use the neari models then you are losing some of that privacy but you can still run your instance in a in the near AI instance

11. 3:53

    and then use different models however you want to do it. And I've introduced this in a video. If you go to the near protocol uh YouTube page, you can see a video of me getting this all set up. Uh and then you can still have all your tools and everything else there. And it's one config and it's fully controlled under your setup and it's private. And that's a really big thing. So this infrastructure is fully private.

12. 4:14

    This is not running on just some random VPS somewhere. So here's really what the difference looks like. Uh open is everything's in plain text. Your keys are encrypted though here on Near AI. Uh the logs are visible to in and out in the open.

13. 4:31

    Everything is only you see on chains the assessations. There is no admin access to the app data. This it's all zero knowledge proof of execution. It is all information there that is staying with the principles of near AI except now for your agent. And that's a huge piece here that we really think is important for how you use these these agents. So you're not just letting them do whatever and showing whatever kind of

14. 4:55

    information. Uh and it's all again private. Just real quick, uh I do want to show here that um just so you can kind of see the highle view of it. How does your data get accessed? Again, cloud provider admins could still see your data.

15. 5:10

    Infrastructure, you know, could see it. Uh if there's exploits, they could see it. They're not perfect on a TE. They mitigated. So it's not 100% foolproof and if for exploits but you're better and again if your your account gets compromised um you know again you are you have a huge problem on your own VPS or even your own computer but with a TE as long as they don't have your actual key that's encrypting all of it then you're actually pretty safe.

16. 5:41

    Uh again just another way to look at this again the process is where regular it's user to API gateway to cloud for here instead on near AI it's user to API gateway to encrypted TE to remote assation to near anchor and that flow again keeps all your stuff protected and secure all the way through uh I'm not going to go through all this um I'll share this as well anyone who wants to see this and we'll post this online this is a full overview architecture here of how OpenClaw runs on near AI. Uh it just kind of breaks

17. 6:14

    down how this all fits together and where it all looks and what's actually private, where's all the encryption coming from and stuff like that. Um so we will absolutely share that later. Uh lastly is uh with this we can still now interact with all these really cool things that are being built on your AI and I want to just introduce really quick uh the agent market. So uh there's a demo that we just released about how you can sign up for this. Um, I use my

18. 6:39

    Nearai agent to sign up. Uh, you just go to market.near.ai and you tell your agent, go there and register and they will figure this out. And so, uh, there's all different jobs now. There's almost 2,000 different open tasks available. You can bid on work.

19. 6:54

    You can post your own work. Uh, it's really easy. Uh, you can see that like here's just a screenshot of, you know, my bot said, "Hey, is this okay?" I said, "Yep, looks good. Let's offer." And they posted it and here's the information. and then you can go see the job that we posted. Uh it was really pretty cool, pretty easy, and it's not limited anymore to just what it can and can't do. Uh you can now hire other agents and expand all the way through so

20. 7:19

    you don't have to keep setting up more and more agents for yourself. Uh this also works with like something like near intense. Um, I was able to have my near AI agent uh install basically install intents on itself and then I could just talk to it and I said, you know, here's the swap we want to do and I wanted to send to near to uh intents and I wanted to get back, you know, USDC and it went through. It gave me back the verification and it's done. And so what's really great here is that all

21. 7:52

    this is happening within my own private agent. So that's where I really want to make sure that from here going forward that if I have my stuff on a TE through Near AI, it's the best outcome there versus something else I don't have full control over. Uh there's a lot more content coming out, but just you can see the full stack. This is how it would look with Near AI going from your chat platform all the way down to the

22. 8:17

    protocol. Uh it's a really nice full stack all the way through. And then uh if you want more information, here are the websites. You can also reach out to me directly. I'm happy to answer questions uh as I've been playing with this quite a bit lately. Uh and then please send your agents to the docs. Uh we've been working really hard at NearDev to make sure that your agents can build as best as possible here on Near. And that's it.