Shade Agents: Truly Autonomous, Non Custodial, Multichain Agents | Matt Lockyer - NEAR & Proximity

[Music] and uh today I'm going to be talking about shade agents the first truly autonomous non-custodial multi-chain agents um you can find out more about shade agents at near. shade so if we look at the state of Agents uh basically in the web two space we obviously have seen some really interesting things happen recently with open ai's operator um and there's a lot of hype just in general around agents in the web 2 space that's pretty you know frothy and Crazy In traditional Tech world as well as in web 3 um so we've seen a lot of uh agentic Frameworks that have launched uh in crypto um and we've also seen some really amazing agents uh for things like signal um and also just entertainment value and digital twinning but we did see some uh some pain points with how these agents are launched um and three specific ones were decentralized Key Management um so basically a lot of agents are run and trusted execution environments today but the keys are not really sort of well managed they're either kind of held by the developers and in The Trusted execution environment if it goes offline maybe the keys lost or it's just a kind of an insecure way to manage keys inside a te itself um and then also there's uh the problem with full code verifiability so how do we know that the agent that's running this code is actually running the code that it says it is uh things like basically upgrade processes and stuff for the agent code um it's all kind of a little bit loose these days like we're not really sure uh what these agents are running and there's no real kind of verifiability there um and also we feel that these agents are lacking smart contract guard rails so basically that final step of sort of protection of for example like user deposits before uh some action can sort of be taken with them um so we set out with a mission to basically enable the building of truly autonomous agents for everyone so we built the shade agent stack um and basically if we look at the top layer of the stack here we have worker agent code so this is any arbitrary code running inside a trusted execution environment so this worker agent code can contact like you know web 2 API storage compute um and then it can bring that back into the trusted execution environment where it's running verifiable code it can also run llms or it could talk to llms outside of the trusted execution environment it can really do kind of like anything that you that you can think of where this is interesting though is that the worker agent is tightly coupled with the smart contract layer that you see at the bottom in the shade agent stack and the worker agent uh basically gets its remote attestation from The Trusted execution environment which basically tells the smart contract that it's running inside a t um but it also gets the code hash of this arbitrary code that it's running so the developer says hey I've got a worker agent here with this particular code and it it runs with this code hash and then the worker agent registers in the smart contract one time and basically says I'm running inside a t smart contract checks that verifies it and then also registered with that with that particular code hash any subsequent calls from the worker agent to the smart contract are basically already PR verified and we can do method Access Control in the smart contracts based on that code hash so this is a very interesting uh kind of part of the the shade agent stack and it opens up a whole new design space so as I mentioned before in current te implementations we've got the keys maybe managed inside the t uh if the T goes offline or if there's some sort of uh you know new code pushed or whatever it's a really insecure process and uh basically we could lose the connection to the protocol and to the users funds and all sorts of things could go wrong so what we did was uh we basically made it so that worker agents don't need to manage Keys um and and basically the worker agent could just go offline and anyone could boot up a new instance of the worker agent provided that they're running the exact same code and still get access to the smart contract and the various methods that manage those keys so that's a key differentiator there pun intended um so this is the registration process for the worker agent code on the left and the smart contract on the right and how you would kind of go all the way through to basically making onchain transactions uh it's a little bit much to go over but you can check this out at near. shade so we've deployed these shade agents already um live on phac cloud so uh we've used them as kind of like a partner for our first shade agent template um and all you really need to do is just dep just deploy some Docker yaml and basically you're up and you're going with your first shade agent in the shade in the worker agent code so in this like arbitrary kind of uh this is Javascript but we also have one that's coming out for the shade template in python as well um there's really just a small amount of code for that agent to basically create their own uh account that can call the smart contract and we create this account with the te's hardware entropy and as I mentioned before the big uh the big kind of differentiator here is that we grab the code has from the te's hardware and we basically uh pass that through to the smart contract register method now this creates a really tight coupling between the code that the worker agent's running and the smart contract so we can do interesting things like method access control so here we have a bunch of worker agent instances we can have many worker agents that all have access to a particular method in a smart contract by virtue of them running that uh specific code that we want them to run we can also have different worker agents operating on different pieces of a protocol so we can have you know worker agent a and worker Agent B operating on different parts of a of a smart contract uh which is really cool we can do this like method method access control so what can you build with shade agents well as we mentioned it's the first truly autonomous agents and in the meme token space I know it's like super hype and froy and stuff um I like to think of it like we see a of agents that are launching with a token but what does that mean you know you got some agent and then you got a token but where's like the the real relationship between them like where's the tight coupling between that token's um you know value Supply or how the token's being managed and like basically how the ecosystem is being managed around the token how is that actually connected to the agent so I like to think of uh shade agents as uh the first way that you can actually launch an agent where the agent is the token so the the agent code that's running inside the trusted execution environment talking to llms and you know maybe doing stuff out on X and stuff like that uh and posting that agent is actually uh fully tightly connected to the token and I'll I'll show you an example of that in just a sec um truly autonomous agents can also deposit and withdraw on any Chain by virtue of using near chain signatures so we can uh do things across any chain I'll also show you an example of that in a sec and we can also handle sensitive user data uh because we're working inside te's which is a confidential Computing environment we can do things like web two payments as far as defi infrastructure is concerned uh for things like oracles we could do things like remove humans in the loop for upgrade processes uh and conflict resolution for web markets I mentioned you know a te can access any web2 apis so we can really just kind of create markets for anything and the last concept I like to think of like it's like the hottest wallet so basically because we're inside a te that can talk to web 2 apis for like oth and other things we can really create these like chain signature wallets that can access you know deposits and stuff on any chain with near chain signatures but we can tightly couple that with uh like an x.com authentication as well so we can basically have users just log in with X and go directly to uh this this hot wallet so as an example agent we used our shade agent template uh and just in in a couple of days we launched this shade agent007 um and I'll tell you what it does in a sec it's just just for fun so uh we basically launched this agent that will uh that will basically read uh tweets on X and it'll basically look for uh terms matching shade agents or AI slop and if the sentiment is positive it'll reward you with base tokens on hyper liquid um or if the sentiment is negative it'll reward you with shade tokens and this is live on near mnet and Hyper liquid mnet and we just we just slapped it together in a couple of days so how this works is basically the worker agent code uh goes out and kind of reads the tweet it analyzes it for the sentiment decides which tokens it's going to give the user and then it creates an evm transaction um and then because the worker agent is running that particular code hash it's registered in the smart contract and it will basically go uh and ask for a signature uh to broadcast on the hyper liquid mainnet from the smart contract and because it runs at code has it will be given the signature and then uh that transaction can be broadcast so uh yeah this is running today uh it's pretty exciting stuff and uh yeah just uh kind of check out shade agents in general and again I just want to hammer home this point of uh method Access Control this is really like the the tight coupling between worker agent code like what is an agent actually doing what is the the brain of an agent actually doing and like how is it kind of like behaving and responding to things and the actual onchain transactions or like parts of the smart contract protocol so thanks for checking out shade agents and once again you can find out more at near. shade