WEBVTT NOTE Sentence-level transcript of https://www.youtube.com/watch?v=gxVZ_1tuuq4 NOTE One cue per sentence. Cue ids are the line anchors on /transcripts/gxVZ_1tuuq4.html. A cue ends where the next begins, or 2 s after its last word. s1 00:00:16.080 --> 00:00:17.600 Hello everybody. s2 00:00:17.600 --> 00:00:19.279 Hello AI engineers. s3 00:00:19.279 --> 00:00:22.400 Are we all having a good time still? s4 00:00:22.800 --> 00:00:23.680 I'm having a good time. s5 00:00:23.680 --> 00:00:24.400 I mean, look at me. s6 00:00:24.400 --> 00:00:24.960 I'm up here. s7 00:00:24.960 --> 00:00:26.400 I'm loving this. s8 00:00:26.400 --> 00:00:28.640 Um, so yeah, thanks so much for joining me. s9 00:00:28.640 --> 00:00:33.120 Um, I want to come and talk to you all about harness engineering and all that kind of stuff. s10 00:00:33.120 --> 00:00:37.040 Um, let me tell you who I am in case you've not met me before. s11 00:00:37.040 --> 00:00:42.239 Um, my name is Mike Chambers and I'm a senior AI specialist developer advocate. s12 00:00:42.239 --> 00:00:45.920 Um, and I work at Amazon at AWS. s13 00:00:45.920 --> 00:00:53.120 Um, little bit about like uh how I managed to get to stand here which is a very exciting time for me. s14 00:00:53.120 --> 00:01:10.479 Um so um quite a while ago in terms of generative AI anyway back in 2023 um I had the amazing awesome privilege uh to work with Antia my colleague at the time and now she works for Amazon AGI you've probably seen her on this stage before s15 00:01:10.479 --> 00:01:20.799 um and the amazing Dr. Andrew Ing on on a course about generative AI with LLMs um sort of can I say that we're approaching half a million enrollments with that? s16 00:01:20.799 --> 00:01:21.920 It looks like that's the case. s17 00:01:21.920 --> 00:01:23.759 And on a three-week course, that's pretty cool. s18 00:01:23.759 --> 00:01:28.159 If you can't tell, in that image, um, I'm playing Transformers with Android. s19 00:01:28.159 --> 00:01:30.880 That seemed like a really funny thing to do at the time. s20 00:01:30.880 --> 00:01:34.159 Um, in 2025, I created an MCP Lambda handler. s21 00:01:34.159 --> 00:01:37.680 It's downloaded still to this day about 35,000 times a month. s22 00:01:37.680 --> 00:01:43.200 Um, to help people in some of the simplest ways of getting serverless MCP serving happening. s23 00:01:43.200 --> 00:01:46.799 Um, I'm going to talk about other things in relation to that this time. s24 00:01:46.799 --> 00:01:48.399 So, we've moved on from that. s25 00:01:48.399 --> 00:01:56.799 Um and in 2026, so the AWS is actually one of the founding members of the Aentka Foundation, part of the Linux Foundation. s26 00:01:56.799 --> 00:01:59.200 Um I'm doing a little bit of work behind the scenes on that. s27 00:01:59.200 --> 00:02:01.119 Hope to do a lot more of that as well. s28 00:02:01.119 --> 00:02:02.880 So little bit about me. s29 00:02:02.880 --> 00:02:14.080 Um so as I've been um preparing for this, oh by the way, I did reread the abstract for this session and realized I said I'd be doing some live coding s30 00:02:14.080 --> 00:02:15.440 and so I will. s31 00:02:15.440 --> 00:02:19.120 So all combined fingers crossed please that that all works for us. s32 00:02:19.120 --> 00:02:28.239 Um but as I've been sort of traveling around a little bit as I do and I was at the AI engineers uh session uh summit uh uh conference in Melbourne s33 00:02:28.239 --> 00:02:32.319 um and took a lot of it in and also from the beginning of this week as well. s34 00:02:32.319 --> 00:02:40.319 I just wanted to to to summarize some of the things that I'm seeing and I'm thinking and I really want to get across and and what really matters to me s35 00:02:40.319 --> 00:02:41.440 and that's this. s36 00:02:41.440 --> 00:02:44.239 There are two different types of agents. s37 00:02:44.239 --> 00:02:49.760 Um, so we talk about agents all the time, but I see two distinct types of agents. s38 00:02:49.760 --> 00:02:55.440 And as I say this, it's going to become really obvious, but they're the agents that we use. s39 00:02:55.440 --> 00:03:00.480 And so, you know, this is claude code and cursor and kirao and all of those types of things. s40 00:03:00.480 --> 00:03:05.519 And also things that don't just generate code, things that we use for productivity and the like as well. s41 00:03:05.519 --> 00:03:08.879 And so those agents we use in a certain type of way. s42 00:03:08.879 --> 00:03:10.800 There's the agents that we use. s43 00:03:10.800 --> 00:03:14.400 And then on the other side of it, we've got the agents that we build. s44 00:03:14.400 --> 00:03:18.959 And that's actually more to do with me and and actually it's more to do with this presentation as well. s45 00:03:18.959 --> 00:03:23.440 It's agents that we build and how we think about agents that we build. s46 00:03:23.440 --> 00:03:29.440 Um and and so I do really think that these don't two things are quite separate and they chain together as well. s47 00:03:29.440 --> 00:03:33.680 I might build an agent that you use and this still holds true. s48 00:03:33.680 --> 00:03:38.720 So token maxing, all that kind of stuff, go for it if that's what you want to do with an agent that you use. s49 00:03:38.720 --> 00:03:42.000 But with an agent that you build, think about it carefully. s50 00:03:42.000 --> 00:03:47.120 Make sure that you're putting it together in a way that's going to work for the audience who's going to use that. s51 00:03:47.120 --> 00:03:50.640 So, I promised that we were talking about harnesses and harness engineering. s52 00:03:50.640 --> 00:03:52.080 So, let's define harness. s53 00:03:52.080 --> 00:03:54.720 I'm sure I'm not the only person to have put something like this up. s54 00:03:54.720 --> 00:03:58.879 I don't usually do this kind of thing and apologize if it makes your skin crawl. s55 00:03:58.879 --> 00:04:01.519 This is a dictionary definition of harness. s56 00:04:01.519 --> 00:04:05.439 A harness is a set of straps and fastenings used to control an animal. s57 00:04:05.439 --> 00:04:12.640 But if we took animal out of here and put model in there, then actually it's pretty right, right, that that's kind of what a harness is. s58 00:04:12.640 --> 00:04:17.919 Other people have done a much better job than just the basic dictionary definition of what a harness is. s59 00:04:17.919 --> 00:04:20.239 So Lang Chain has got a article out. s60 00:04:20.239 --> 00:04:21.519 You've probably seen stuff like that. s61 00:04:21.519 --> 00:04:24.160 Martinfowler.com although Martin didn't write it. s62 00:04:24.160 --> 00:04:25.520 It was Pitta wrote this. s63 00:04:25.520 --> 00:04:28.240 Harness engineering for coding agents. s64 00:04:28.240 --> 00:04:29.840 Agents that we use, right? s65 00:04:29.840 --> 00:04:32.320 So there are other ways of looking at harnesses. s66 00:04:32.320 --> 00:04:33.759 What about Amazon then? s67 00:04:33.759 --> 00:04:36.320 How do we see um harnesses? s68 00:04:36.320 --> 00:04:39.120 Well, um no, okay, this is the wrong kind of harness. s69 00:04:39.120 --> 00:04:41.360 Sorry, we do have strong opinions on harnesses. s70 00:04:41.360 --> 00:04:44.720 I'm going to show you all of that, but we sell all kinds of things. s71 00:04:44.720 --> 00:04:55.600 So, in a nutshell, and if you read those articles, um and if you've had the conversations around here at this event, of course, um a harness, you take an agent, s72 00:04:55.600 --> 00:05:00.400 remove the model part from it, and everything that you have left, that's the harness. s73 00:05:00.400 --> 00:05:06.160 Okay, so let's think about that in context of an agent that we use. s74 00:05:06.160 --> 00:05:09.199 And so it's pretty, I think, fairly straightforward. s75 00:05:09.199 --> 00:05:10.720 I have this coding assistant. s76 00:05:10.720 --> 00:05:11.680 It's probably on my machine. s77 00:05:11.680 --> 00:05:13.600 It has access to my files. s78 00:05:13.600 --> 00:05:28.720 And I create a harness or the place I work at has created a harness for me which contains um how it's going to use memory, the skills that I want it to use, tools and MCP servers to allow it to go to be able to go and connect to documentation s79 00:05:28.720 --> 00:05:29.840 servers and the like. s80 00:05:29.840 --> 00:05:35.199 and and well set up engineering teams have got their standards that they've had. s81 00:05:35.199 --> 00:05:44.000 They've had coding standards for decades, but now they have basically harness standards, the things that they want to deploy to everybody's coding assistance. s82 00:05:44.000 --> 00:05:46.000 So, in a nutshell, that's what it is. s83 00:05:46.000 --> 00:05:48.240 I'm not going to talk to too much more about that. s84 00:05:48.240 --> 00:05:53.919 Um, but I do want to share one QR code with you, and I'll try and give you a little bit of warning before I bring QR codes out. s85 00:05:53.919 --> 00:05:56.720 This is the agent toolkit for AWS. s86 00:05:56.720 --> 00:05:58.639 This is u available on GitHub. s87 00:05:58.639 --> 00:05:59.840 Of course, it's free. s88 00:05:59.840 --> 00:06:04.400 You can install it and it helps you if this is what you're doing and you're deploying code. s89 00:06:04.400 --> 00:06:15.520 If you're deploying on AWS or you're thinking about deploying on AWS or or maybe you will one day, grab this toolkit, enable your um uh agent to to help you in the right direction. s90 00:06:15.520 --> 00:06:18.560 It's instructions for how to install it on pretty much everything. s91 00:06:18.560 --> 00:06:24.479 And the reason why I get passionate about this is because I don't want to see any more slop ops. s92 00:06:24.479 --> 00:06:30.560 Um, so we always used to push back against click ops in, you know, in the professional cloud development space. s93 00:06:30.560 --> 00:06:36.960 Clicking around on the console is great for being able to figure out what's going on, but it's not how you deploy things into production. s94 00:06:36.960 --> 00:06:45.440 We can ask an agent what's going on, but we don't want to ask the agent to spin up an S3 bucket, get me an EC2 instance, whatever it might be. s95 00:06:45.440 --> 00:06:54.240 We want the agent to build up our infrastructure as code which is going to go and do that so that we still own our deployments in the cloud. s96 00:06:54.240 --> 00:06:56.560 So no more swap ops. s97 00:06:56.560 --> 00:06:56.880 Okay. s98 00:06:56.880 --> 00:06:58.560 So that's the agent that we use. s99 00:06:58.560 --> 00:07:00.960 Now let's go and talk about the agent that we're going to build. s100 00:07:00.960 --> 00:07:06.160 And I'm going to get into the code as quickly as I can and we'll do as much as it has time for. s101 00:07:06.160 --> 00:07:10.800 So how do we think about a harness in relation to the agent that we're building? s102 00:07:10.800 --> 00:07:12.880 Exactly the same to a point. s103 00:07:12.880 --> 00:07:13.440 Yes. s104 00:07:13.440 --> 00:07:16.160 We still want to have how are we going to manage the memory? s105 00:07:16.160 --> 00:07:22.400 How are we going to manage the skills and tools in MCP by the way that that belies a lot of stuff, right? s106 00:07:22.400 --> 00:07:30.240 Because you can pretty much extend an agent to do almost anything you want with a whole bunch of different types of tools which could be via MCP. s107 00:07:30.240 --> 00:07:36.240 But with an agent that I am building, I need to think about a lot more than just that. s108 00:07:36.240 --> 00:07:43.680 especially if um like at Amazon and like at cloud scale I'm deploying my agent out to the masses. s109 00:07:43.680 --> 00:07:46.639 So how do I actually manage the loop? s110 00:07:46.639 --> 00:07:56.479 How do I manage um scaling payments, memory, identity, skills, runtime, context management, the rest of it and I have left it to the last thing but it should be the first thing that I say. s111 00:07:56.479 --> 00:08:00.720 Observability and evaluations super super important. s112 00:08:00.720 --> 00:08:02.560 How do we actually deal with this? s113 00:08:02.560 --> 00:08:07.440 Do I write all of this code down into one container and just deploy it and scale that? s114 00:08:07.440 --> 00:08:08.160 Not really. s115 00:08:08.160 --> 00:08:15.680 If I want to be scaling to thousands of users, I need to think about each individual of these components and how I'm going to scale them out individually. s116 00:08:15.680 --> 00:08:19.360 And that to me is harness engineering. s117 00:08:19.360 --> 00:08:21.039 This is the serious side of stuff. s118 00:08:21.039 --> 00:08:25.199 This is the big stuff that we want to get harnesses working at real scale. s119 00:08:25.199 --> 00:08:27.440 Okay, let's see if this works. s120 00:08:27.440 --> 00:08:33.519 I can feel your combined goodwill being sent my way that we're going to try and make some code work. s121 00:08:33.519 --> 00:08:34.640 So, I'm here in Kira. s122 00:08:34.640 --> 00:08:36.399 This is my IDE of choice here. s123 00:08:36.399 --> 00:08:42.479 And I've got a few different samples that we're just going to race through watching that clock countdown fast. s124 00:08:42.479 --> 00:08:48.320 So, um just just make sure that we're all on the same page here and hopefully you can all see this. s125 00:08:48.320 --> 00:08:55.040 Um of the code which I'm about to show you, by the way, one piece of code, not this one, has been generated by Curo. s126 00:08:55.040 --> 00:09:00.080 Everything else is either a tool or this one I actually wrote it myself. s127 00:09:00.080 --> 00:09:01.600 I didn't use an agent for this. s128 00:09:01.600 --> 00:09:04.966 I know I deserve a round of applause, but it's okay. s129 00:09:04.966 --> 00:09:05.839 [laughter] s130 00:09:05.839 --> 00:09:09.279 So, this is uh this is a Strand agent. s131 00:09:09.279 --> 00:09:11.680 So, I've just taken the Strands agents SDK. s132 00:09:11.680 --> 00:09:14.000 Um, and hopefully this kind of thing is kind of familiar. s133 00:09:14.000 --> 00:09:15.200 I've brought in an agent. s134 00:09:15.200 --> 00:09:18.959 I brought in the tool decorator and I'm creating myself an agent. s135 00:09:18.959 --> 00:09:20.959 The tool definition is down here. s136 00:09:20.959 --> 00:09:23.680 Um, and so I just pass in my system prompt. s137 00:09:23.680 --> 00:09:27.360 Things are pretty simple at this stage and I've passed in a couple of tools. s138 00:09:27.360 --> 00:09:37.040 Calculator is something that's a library I can install and get time is the one that we always use because I don't tend to use agents to book flights certainly not ones like this. s139 00:09:37.040 --> 00:09:40.240 Um and so I can say something simple here like what is the time? s140 00:09:40.240 --> 00:09:44.640 I'm not going to run this because you know the time but you can see generally how this works. s141 00:09:44.640 --> 00:09:46.720 Is this a harness? s142 00:09:46.720 --> 00:09:47.519 Sort of. s143 00:09:47.519 --> 00:09:49.120 There's not an awful lot to it, right? s144 00:09:49.120 --> 00:09:50.399 We've got the tools in there. s145 00:09:50.399 --> 00:09:53.360 Our loop is being managed for us by the framework. s146 00:09:53.360 --> 00:09:54.160 This is pretty cool. s147 00:09:54.160 --> 00:09:55.040 So that's good. s148 00:09:55.040 --> 00:09:58.560 But obviously if I was to run this, this is running on my laptop. s149 00:09:58.560 --> 00:10:00.800 It's not running at any particular scale. s150 00:10:00.800 --> 00:10:05.360 And we're missing some of the attributes that I want from the agents that I'm going to deploy. s151 00:10:05.360 --> 00:10:08.320 Let me move on to my next agent quickly. s152 00:10:08.320 --> 00:10:17.360 So this is also a strand agent, but this one I actually asked Kira to write it for me for this session um because I wanted to include some more stuff. s153 00:10:17.360 --> 00:10:25.360 And so inside of this agent, the one main thing that I want to point out is that I am uh included a session manager. s154 00:10:25.360 --> 00:10:30.800 So my session manager is helping me to maintain session state between invocations. s155 00:10:30.800 --> 00:10:32.800 So this is a sort of memory. s156 00:10:32.800 --> 00:10:35.760 It's a kind of medium-term short-term memory kind of thing. s157 00:10:35.760 --> 00:10:38.560 It's not proper long-term memory, but it is there. s158 00:10:38.560 --> 00:10:45.360 And actually, it does store long-term memories in files which are down the side here that it's uh included for us. s159 00:10:45.360 --> 00:10:50.959 So, if I just scroll down here, you can see uh yeah, here's the agent definition itself. s160 00:10:50.959 --> 00:10:54.959 Um, and we've got a bit more of a system prompt because Kira couldn't help itself. s161 00:10:54.959 --> 00:10:57.600 Um, and we've got some tools here defined. s162 00:10:57.600 --> 00:11:02.720 Um, including a remember tool that the agent can decide to use to remember stuff about me. s163 00:11:02.720 --> 00:11:05.680 Um, and then I've got my uh session manager down there. s164 00:11:05.680 --> 00:11:12.480 And that session manager is going to rehydrate the conversation history when I come back to chat to it the next time. s165 00:11:12.480 --> 00:11:15.120 and maybe the next time is now. s166 00:11:15.120 --> 00:11:16.640 So, let's see if we can get this working. s167 00:11:16.640 --> 00:11:18.880 Now, again, this is running on my local machine. s168 00:11:18.880 --> 00:11:20.880 Um, and this is a demo here. s169 00:11:20.880 --> 00:11:25.279 So, let's just type in hello because I'm scared of typing too much and spelling it wrong. s170 00:11:25.279 --> 00:11:27.440 Um, and it says, uh, yeah, keep testing me. s171 00:11:27.440 --> 00:11:28.079 Bring it on. s172 00:11:28.079 --> 00:11:28.720 Excellent. s173 00:11:28.720 --> 00:11:35.440 Um, um, who will win the World Cup? s174 00:11:35.440 --> 00:11:37.760 So, obviously I need to know this. s175 00:11:37.760 --> 00:11:40.320 And, um, what does it say? s176 00:11:40.320 --> 00:11:40.640 Yeah. s177 00:11:40.640 --> 00:11:44.000 So, while you could just say Australia, it knows I want Australia to win. s178 00:11:44.000 --> 00:11:45.440 It's where I'm currently living. s179 00:11:45.440 --> 00:11:46.720 I'm Australia. s180 00:11:46.720 --> 00:11:51.200 So, obviously Australia is going to win the World Cup, but why has it got that? s181 00:11:51.200 --> 00:11:53.839 It's because of previous conversations that we've had. s182 00:11:53.839 --> 00:12:00.800 And obviously, it's being honest that it has no clue because that's coming from the large language model, of course. s183 00:12:00.800 --> 00:12:04.800 So, okay, looks at a couple of different agents there, blasted through this. s184 00:12:04.800 --> 00:12:07.360 This is um running on my machine. s185 00:12:07.360 --> 00:12:15.040 So this isn't really getting me to clouds scale of course and I'm I'm picking up and I'm including various pieces in this like memory. s186 00:12:15.040 --> 00:12:16.560 So let's go next. s187 00:12:16.560 --> 00:12:29.040 How do we get to the point where I can deploy something like this if not this actual agent out at cloudscale and take things like memory and deploy that separately so it can scale separately. s188 00:12:29.040 --> 00:12:35.600 taking our loop out so it can scale separately and we can then bolt in all kinds of other things as well. s189 00:12:35.600 --> 00:12:42.240 So in order to do that I'm going to use something called um agent core um and so we have bedrock agent core. s190 00:12:42.240 --> 00:12:47.040 It's part of the stack that we have at AWS and that's how I'm doing this and how I'm deploying. s191 00:12:47.040 --> 00:12:52.800 So I've done that already but I want to show you how to start out with that and how we do this. s192 00:12:52.800 --> 00:12:56.880 So if I go to here uh yeah I'm ready to go. s193 00:12:56.880 --> 00:13:01.279 So, I have a command line tool on my machine, the agent core command line. s194 00:13:01.279 --> 00:13:05.680 Um, and so there's a QR code at the end, as you might imagine, so that you can get hold of this. s195 00:13:05.680 --> 00:13:09.360 Um, but I can use this to help me deploy my agent. s196 00:13:09.360 --> 00:13:12.880 Now, this steps me through like many of these types of tools do. s197 00:13:12.880 --> 00:13:15.200 Um, and it sort of steps me through what do you want to do? s198 00:13:15.200 --> 00:13:17.440 So, this is my woohoo agent. s199 00:13:17.440 --> 00:13:19.600 Um, and it's going to ask me a bunch of stuff. s200 00:13:19.600 --> 00:13:22.240 And I wanted to show you some of this as we step through. s201 00:13:22.240 --> 00:13:25.200 Now, strangely, I'm not going to select harness. s202 00:13:25.200 --> 00:13:29.920 And we'll come back to why I'm not selecting harness in a second, but I'm saying I wanted to deploy an agent. s203 00:13:29.920 --> 00:13:36.959 And what's going to happen here is this command line tool is actually going to step me through and actually write an entire agent. s204 00:13:36.959 --> 00:13:41.279 It's basically a hello world agent that I can then go and customize myself. s205 00:13:41.279 --> 00:13:45.839 Um, and so using this command line is an easy way to get started with agent core. s206 00:13:45.839 --> 00:13:47.360 So I'm going to keep the default name. s207 00:13:47.360 --> 00:13:51.040 In fact, I'm probably going to keep all the defaults here just so we can see what's the option. s208 00:13:51.040 --> 00:13:55.920 Of course, I can bring code if I want, but I'm going to ask it to create some code for me. s209 00:13:55.920 --> 00:13:57.279 So, it says, well, what do you want? s210 00:13:57.279 --> 00:13:58.800 Python or TypeScript? s211 00:13:58.800 --> 00:14:04.399 And back in the day, I used to do things like activate functions and back propagation in the machine learning space. s212 00:14:04.399 --> 00:14:06.720 So, Python it is for me. s213 00:14:06.720 --> 00:14:08.160 So, I will choose that. s214 00:14:08.160 --> 00:14:10.480 Um, and there's some deployment options. s215 00:14:10.480 --> 00:14:15.120 There's also this I just want to point this out like how can we actually go and connect into our agent. s216 00:14:15.120 --> 00:14:22.880 So, our agent that's running at scale in the cloud, HTTP is probably the obvious one, but we might want to have it being served behind MCP. s217 00:14:22.880 --> 00:14:30.959 We might want to use awesome little things like AGUI so we can make nice interactive chat agents, but I'm going to say HTTP. s218 00:14:30.959 --> 00:14:34.000 We can use any um framework we want. s219 00:14:34.000 --> 00:14:42.399 I happen to use Strand's agents SDK, but anything you could write your own framework if you want to um or your own own base code. s220 00:14:42.399 --> 00:14:44.720 Any model is supported by this as well. s221 00:14:44.720 --> 00:14:47.199 So, we don't just have to use the Amazon models. s222 00:14:47.199 --> 00:14:51.199 Um, and we don't have to use the ones from Amazon Bedrock, but we can use any model. s223 00:14:51.199 --> 00:14:52.079 I'm using the one here. s224 00:14:52.079 --> 00:14:56.320 I'm using Sonic 4.5 just because that's offered to me at default. s225 00:14:56.320 --> 00:14:57.279 And here's memory. s226 00:14:57.279 --> 00:14:58.880 So, this is the one thing I wanted to show you. s227 00:14:58.880 --> 00:15:02.720 So, I can come in here and ask for long-term and short-term memory to be deployed. s228 00:15:02.720 --> 00:15:11.839 And we'll see what this means in just a second, but it's basically going to create for us cloud infrastructure which is going to manage those memories for us separately s229 00:15:11.839 --> 00:15:16.639 from our running agent running asynchronously from our agent and connected of course. s230 00:15:16.639 --> 00:15:18.800 So there's obviously other kinds of things we can do. s231 00:15:18.800 --> 00:15:23.600 We can hit enter and it will start to create the configuration of this agent on my machine. s232 00:15:23.600 --> 00:15:30.720 Now I'm going to skip over here and come back to the actual code I have because I've already done this of course. s233 00:15:30.720 --> 00:15:33.839 Um, and this is the agent that it would be currently deploying. s234 00:15:33.839 --> 00:15:34.880 Something like this. s235 00:15:34.880 --> 00:15:36.240 So, we've built up here. s236 00:15:36.240 --> 00:15:37.760 This is a little bit more complex. s237 00:15:37.760 --> 00:15:39.920 So, this is a strands agent. s238 00:15:39.920 --> 00:15:42.800 You'll notice that it's got a few more things added in. s239 00:15:42.800 --> 00:15:52.959 So, it's got the linkage into Amazon Bedrock agent core app, but pretty much apart from that, that's all you need in order to be able to scale this agent out at runtime s240 00:15:52.959 --> 00:15:55.600 and do multi-tenant isolation. s241 00:15:55.600 --> 00:16:03.759 So you can write an agent that works for one user and then scale that out without you having to write all the multi-tenented code. s242 00:16:03.759 --> 00:16:09.839 It's a massive saver and from a security and identity perspective, it's makes it so much simpler. s243 00:16:09.839 --> 00:16:12.000 It's um it's very very useful. s244 00:16:12.000 --> 00:16:15.839 So if I scroll down through here, you can see the rest of it is looking pretty similar. s245 00:16:15.839 --> 00:16:17.759 We've got some test tools in here. s246 00:16:17.759 --> 00:16:21.199 We've got a connection to MCP, so we can see how that is done. s247 00:16:21.199 --> 00:16:28.240 Um, and we've got the uh connection into our session manager and our memory which is all built in here. s248 00:16:28.240 --> 00:16:34.800 So, if I scroll down a bit more, we'll be able to see somewhere where we actually invoke the thing um which is there. s249 00:16:34.800 --> 00:16:37.839 Um, and the system prompt is is somewhere at the top. s250 00:16:37.839 --> 00:16:39.839 So, we can we can scroll through this code. s251 00:16:39.839 --> 00:16:40.959 I'm going through it quickly. s252 00:16:40.959 --> 00:16:43.519 You can write your own code and do this with it as well. s253 00:16:43.519 --> 00:16:47.839 If I go back over to my um uh code here for just one second. s254 00:16:47.839 --> 00:16:55.680 I'm in the folder now that has been created with that code locally, it is actually deployed, but let's assume it's not deployed quite yet. s255 00:16:55.680 --> 00:16:58.639 I can come back in and type in agentcore dev. s256 00:16:58.639 --> 00:17:05.760 And what that's going to do for me um Wi-Fi permitting is it will spin up for us a web uh browser. s257 00:17:05.760 --> 00:17:09.600 And inside of that web browser, we're now connected to that agent running locally. s258 00:17:09.600 --> 00:17:14.000 So if I make updates to that code, we would see that happen in real time here. s259 00:17:14.000 --> 00:17:20.400 So I can say hello, I am doing the press now. s260 00:17:20.400 --> 00:17:23.679 Um it knows that I'm coming to do a presentation, but I think it does anyway. s261 00:17:23.679 --> 00:17:26.480 Um and so yeah, you can interact with the agent here. s262 00:17:26.480 --> 00:17:31.120 You can make adjustments to the um to the code and you'll see it update live here. s263 00:17:31.120 --> 00:17:36.640 Um but you can also use this to switch over to the live um deployed version. s264 00:17:36.640 --> 00:17:54.400 So with agent core deploy it will use infrastructure as co code like I talked about before to deploy your agent out at scale with the memory with the agent with runtime and with many other components if you choose to do so you can use this interface then to go and look at traces look at memory stored look s265 00:17:54.400 --> 00:17:58.080 at all that stuff so that you can debug and see what's going on. s266 00:17:58.080 --> 00:18:06.559 Now, when we stepped through the um the the the console just a second ago through the the CLI app a second ago, we didn't choose harness. s267 00:18:06.559 --> 00:18:10.480 I skipped out on that one and I'm just going to show you that quickly now. s268 00:18:10.480 --> 00:18:16.320 So, one thing we can do instead is I think we can get to the point you've seen I've deployed agents. s269 00:18:16.320 --> 00:18:17.200 I didn't do very much. s270 00:18:17.200 --> 00:18:20.160 I just did a system prompt and some tools and go. s271 00:18:20.160 --> 00:18:29.520 And there's an argument to be made that essentially if that's possible then maybe 80% of um agentic use cases 80% of agent development is kind of solved already. s272 00:18:29.520 --> 00:18:34.799 We don't need to do much more than system prompt connect to some MCP tools and we've got what we want. s273 00:18:34.799 --> 00:18:39.280 And if that's the case then we have harness built into agent core. s274 00:18:39.280 --> 00:18:41.679 This is the configuration for an agent. s275 00:18:41.679 --> 00:18:48.640 I just have a simple JSON which is showing me which model do I want to use and what system prompt do I want to use. s276 00:18:48.640 --> 00:18:51.039 Can't get much more simpler than that system prompt. s277 00:18:51.039 --> 00:18:54.880 Um, and then this can also be deployed with agent core deploy. s278 00:18:54.880 --> 00:18:58.960 So at this point we don't have even any agentic code either. s279 00:18:58.960 --> 00:19:00.880 We can just deploy it straight out. s280 00:19:00.880 --> 00:19:06.160 If you want to know any more about any of this then please do come and see us down on the booth or see me after this session. s281 00:19:06.160 --> 00:19:09.440 I'll be more than happy to talk to you at length about this. s282 00:19:09.440 --> 00:19:14.160 18 minutes is such a short amount of time for me to be able to talk about almost anything. s283 00:19:14.160 --> 00:19:18.480 But this is essentially all of these components on here. s284 00:19:18.480 --> 00:19:22.080 There's a mapping somewhere into something that Agent Core has. s285 00:19:22.080 --> 00:19:23.360 I apologize for the colors. s286 00:19:23.360 --> 00:19:24.880 It seemed like a good idea at the time. s287 00:19:24.880 --> 00:19:32.400 Um, but this is an overview of the different capabilities that are composable out of agent core. s288 00:19:32.400 --> 00:19:38.000 So, you can take any of these and use any of them together or separately. s289 00:19:38.000 --> 00:19:49.039 If you have an agent that's running in production very happily at the moment, but you like the idea of having long-term memory managed for you serverless, then you can just take that part and integrate it. s290 00:19:49.039 --> 00:19:50.320 That's totally something you can do. s291 00:19:50.320 --> 00:19:51.200 Here's a QR code. s292 00:19:51.200 --> 00:19:53.360 Sorry, probably should have put that there a second ago. s293 00:19:53.360 --> 00:20:05.520 I'm moving this QR code in just a moment, but Amazon Bedrock agent core is that if you're interested in the Strands agents, which I happen to be using for this, it's obviously it's it's free because it's open source. s294 00:20:05.520 --> 00:20:10.000 Um, and it's a um a model first framework for putting together agents. s295 00:20:10.000 --> 00:20:15.120 It's super fast, it's super powerful, and it's what I use all the time. s296 00:20:15.120 --> 00:20:17.840 Thank you so much for being with me in this presentation. s297 00:20:17.840 --> 00:20:20.000 Please feel free to connect with me on LinkedIn. s298 00:20:20.000 --> 00:20:22.640 I'd love to carry on the conversation with you. s299 00:20:22.640 --> 00:20:28.480 have a fantastic rest of show and have a safe travel as you go home after the event. s300 00:20:28.480 --> 00:20:30.960 Thank you so much.