YouTube2Text

Agentforce NOW codeLive: Build Agentic Data 360 Experiences Headlessly — Transcript

by Salesforce Developers · 11,682 words · 1,787 segments · language en · Watch on YouTube

Full transcript

  1. 0:00Hello everyone. Thank you so much for
  2. 0:02joining Agent Force now Code Live. My
  3. 0:06name is Mohit. I'm a developer advocate
  4. 0:08here at Salesforce and I welcome you all
  5. 0:11to this Code Live episode. Today we're
  6. 0:13going to be
  7. 0:15building agentic data experiences, more
  8. 0:18importantly headlessly, all through
  9. 0:20coding agents.
  10. 0:22Uh and I have with me uh the product
  11. 0:25manager who is sort of working on all of
  12. 0:28this. Anush, welcome to Code Live.
  13. 0:31>> Yeah, thank you Mohit. Thank you for
  14. 0:32inviting me.
  15. 0:33Uh hi everyone. My name is Anush. I'm
  16. 0:36the product manager for headless data
  17. 0:37360 starting with the data 360 MCP
  18. 0:41server.
  19. 0:42So I'm really excited to put this
  20. 0:43hands-on session with Mohit to get you
  21. 0:45guys started and to see your experience
  22. 0:47using the MCP server.
  23. 0:49>> Perfect. Perfect. And and I know Anush
  24. 0:51you're very busy building these MCP
  25. 0:53servers, so I really appreciate you
  26. 0:56taking time today. And for all those of
  27. 0:58you who are tuning
  28. 1:00um remember that we also have chats
  29. 1:02there. So as we sort of show these
  30. 1:04capabilities to you, if you have any
  31. 1:07questions, please make use of those chat
  32. 1:11and um shoot us to us and we'll we'll
  33. 1:13get back to those questions as they sort
  34. 1:15of come along. Uh with that uh you know,
  35. 1:18because this is Salesforce presentation,
  36. 1:21um
  37. 1:22I'll put the presentations here, but
  38. 1:24more importantly uh our forward-looking
  39. 1:26statement. Uh so please do not make your
  40. 1:29purchasing decisions based on features
  41. 1:33um that are uh not available in the
  42. 1:35product yet or not GA.
  43. 1:37With that out of our way, again um you
  44. 1:40know, myself and Anush we're going to be
  45. 1:41talking about building agentic data 360
  46. 1:45experiences headlessly.
  47. 1:47And I just want to remind everyone that
  48. 1:50uh you know, there is a QR code. I've
  49. 1:52also uh sort of placed this QR code on
  50. 1:54your screens on your right. So um
  51. 1:57make sure that if you have any feedback
  52. 1:59about this code live or any code live in
  53. 2:02general, just use that QR code and tell
  54. 2:06us. With that,
  55. 2:08how about Anush, we actually take some
  56. 2:11time and explain
  57. 2:13to everyone
  58. 2:15like what is this data 360 headless all
  59. 2:17about, what is this momentum about, what
  60. 2:19are these MCP servers. So, let's take
  61. 2:21some time everyone to get to the same
  62. 2:23page here.
  63. 2:24>> Yes.
  64. 2:25I think that'd be a good start.
  65. 2:27Starting with what are we doing here?
  66. 2:29If you started earlier this year using
  67. 2:31data 360, maybe even in Jan,
  68. 2:34a lot of the functionality was in the
  69. 2:37lightning UI in Salesforce.
  70. 2:39So, even if you're using agents back
  71. 2:41then, you had to move between the UI,
  72. 2:44maybe ask a coding agent what you need
  73. 2:46to do, look at the documentation. So,
  74. 2:48this led to like some sort of context
  75. 2:50mismatch between where you're doing work
  76. 2:52and where you need to go gather this
  77. 2:53context for.
  78. 2:55So, in order to solve that, we are
  79. 2:58starting a really focused headless
  80. 3:00program.
  81. 3:02What does this mean?
  82. 3:04The foundation of this program for us is
  83. 3:06making sure all our functionality in the
  84. 3:09UI is available through APIs. So, that
  85. 3:11is the foundation layer in which we are
  86. 3:14building.
  87. 3:15Once those APIs are exposed, we have now
  88. 3:18released the data 360 MCP server.
  89. 3:22Actually, earlier in June we had a
  90. 3:23developer preview of a local MCP server
  91. 3:26>> Right.
  92. 3:26>> we got a lot of feedback about how our
  93. 3:28users are using it. We saw the user
  94. 3:31experience with a lot of great feedback
  95. 3:33from that.
  96. 3:34And from that we realized that the
  97. 3:36hosted remote MCP server was a better
  98. 3:38option for most users.
  99. 3:41>> Perfect.
  100. 3:42>> Now,
  101. 3:44why is this really important, right?
  102. 3:47It's not just about you going to Claude
  103. 3:49or going to Codex and telling uh
  104. 3:51Codex, go do this thing or Claude do
  105. 3:53this thing no make no mistakes. It's not
  106. 3:55just that. It's about providing trusted
  107. 3:57context
  108. 3:59and also with governed resources and
  109. 4:03pre-built skills via the MCP server.
  110. 4:06Why is this important?
  111. 4:09So, if you notice
  112. 4:10agents
  113. 4:12are good at solving things but they may
  114. 4:15also not be the best at following the
  115. 4:17shortest path to a solution.
  116. 4:20But you being Salesforce developers or
  117. 4:22industry experts know what is the best
  118. 4:24path.
  119. 4:25So, this MCP server paired with
  120. 4:28specialized skills reduce the number of
  121. 4:31tokens required for an agent to go from
  122. 4:33zero to one.
  123. 4:34So if you notice the MCP server today,
  124. 4:37we have just three tools.
  125. 4:39We have the search tool.
  126. 4:42We have payload examples and we have
  127. 4:44execute.
  128. 4:46Why just three tools, right?
  129. 4:49We notice that with dynamic tool
  130. 4:51loading, when you ask for a certain
  131. 4:54intent the agent can then then search
  132. 4:57for what what it needs. For example, it
  133. 4:59could be segmentation. You want to
  134. 5:01create a user segment to maybe
  135. 5:03uh
  136. 5:04target a specific user segment. And then
  137. 5:07the agent finds the exact payload
  138. 5:08examples for that tool. And that reduces
  139. 5:11the number of tones an agent needs
  140. 5:14to successfully solve this.
  141. 5:16So, which is why we have followed this
  142. 5:17dynamic tool loading method as well as
  143. 5:20coming soon will be skills accompanied
  144. 5:23with the MCP server.
  145. 5:24This will reduce the number of tones an
  146. 5:26agent needs, the number of tokens you
  147. 5:28need, and also it provides best
  148. 5:30practices that practitioners using the
  149. 5:33Data 360 platform used to build it.
  150. 5:36S-
  151. 5:38So now, not just can you do this through
  152. 5:40Claude or Codex, but you get this really
  153. 5:42smart reasoning engine paired with you
  154. 5:45following the best practices and
  155. 5:47executing on it. Which is why I'm so
  156. 5:48excited for this.
  157. 5:50>> Awesome. Awesome. No, this is this is
  158. 5:52great. Um one thing I really want to
  159. 5:54call out is look at the simplicity of
  160. 5:57this whole design, right? Uh when I when
  161. 6:00I was sort of um
  162. 6:02thinking about, you know, these MCPs for
  163. 6:05Data360 and myself, you know, um
  164. 6:08I'm I I can call myself Data360 uh
  165. 6:12developer or
  166. 6:13>> you can.
  167. 6:14>> you know, Data Blazer. Um
  168. 6:16so, there are there are so many tools in
  169. 6:19Data360, as as you know, um Anush, like
  170. 6:23there's so much of power that this
  171. 6:25Data360 has and the tools it has. You
  172. 6:28know, let's talk about, you know,
  173. 6:29creating these data streams to bring
  174. 6:31data from like, you know, multiple
  175. 6:33system, like thousands of systems,
  176. 6:35right? You can ingest data through
  177. 6:37various mechanisms, like, for example,
  178. 6:39zero copy is one of my favorite. Because
  179. 6:42if you have a Snowflake
  180. 6:43>> Yeah.
  181. 6:45>> you know, sort of zero copy uh data into
  182. 6:47Data360.
  183. 6:49Uh but still like all of those involve a
  184. 6:51lot of clicks. And when I thought about
  185. 6:53MCPs for, you know, Data360
  186. 6:56with all these tools, I was like, this
  187. 6:58is going to be like, you know, maybe n
  188. 7:00number of tools, right? Thousands of
  189. 7:02tools here in this
  190. 7:03>> Yeah.
  191. 7:03>> in this scenario because there are
  192. 7:05thousands of APIs. But uh you know, when
  193. 7:07I saw like, okay, you know, you guys are
  194. 7:09like only exposing like only three
  195. 7:12different MCP tools,
  196. 7:14um so, that's again less number of token
  197. 7:17consumptions and also much easier for
  198. 7:20the AI to like sort of reason, right?
  199. 7:22Because it's now progressively sort of
  200. 7:24progressing towards what what is
  201. 7:26required than, you know, just loading
  202. 7:29everything in one shot and sort of
  203. 7:31consuming the context window and then
  204. 7:33sort of having AI just hallucinate it,
  205. 7:35right?
  206. 7:35>> Yeah.
  207. 7:37>> Perfect. Perfect. Okay. Uh so, for all
  208. 7:39of you, right? Um today what I what what
  209. 7:41we are thinking is we're going to make
  210. 7:43this much more like practical just
  211. 7:46showing the use of it. Um, and uh if you
  212. 7:49are new to Data 360, welcome. Um, no
  213. 7:52issues. We're we'll try to like cover
  214. 7:54those aspects as well. Like what is Data
  215. 7:56360? What does it do? Etc.
  216. 7:59Um, but I have a business use case and I
  217. 8:02think, you know, all of us are there to
  218. 8:04solve business problems, right? So, I
  219. 8:06thought it will resonate well if we
  220. 8:08start with the the use case here.
  221. 8:10And uh Anush, for us today the use case
  222. 8:13is we have this Pronto, a food delivery
  223. 8:15platform. And this is a sample
  224. 8:17application that developer advocacy team
  225. 8:19sort of uses it uh to showcase uh
  226. 8:22different capabilities of the platform.
  227. 8:25So, we are taking that Pronto uh
  228. 8:27platform today and we will leverage the
  229. 8:29Data 360 as a centralized system for
  230. 8:32context for agents that Pronto is
  231. 8:35building. Um, basically it's going to be
  232. 8:38unifying data from multiple sources,
  233. 8:41activating that data across different
  234. 8:43applications and different agents. So,
  235. 8:45this is the primary use case of Data 360
  236. 8:48in case you are wondering what is Data
  237. 8:49360, right? Is that context engine that
  238. 8:52will drive agents, uh your apps, you
  239. 8:55know, think of it like a big data
  240. 8:57processing engine, right? With all the
  241. 8:58tools and batteries included so you can
  242. 9:01quickly activate, you know, the data um
  243. 9:04for context as well as for um sort of
  244. 9:08building applications if if that's what
  245. 9:10you're doing with this data, right? Um,
  246. 9:12so that's so our use case, our excite
  247. 9:14business problem that we're going to
  248. 9:16solve we're going to try to solve in
  249. 9:17this session. Although, I don't know how
  250. 9:19far we will get along with these coding
  251. 9:21agents and sort of working because, you
  252. 9:23know, building something live is like,
  253. 9:26you know, hours of work. Like even
  254. 9:27clicking through it. And agents is going
  255. 9:29to simplify a lot of it. We'll try to to
  256. 9:31sort of do a lot of stuff here um live.
  257. 9:34So, really what we are trying to solve
  258. 9:36is uh for Pronto, the customer is
  259. 9:39showing up now in as two different
  260. 9:41peoples. Even though it's same customer,
  261. 9:44um but it's showing up as two different
  262. 9:46people because uh you know,
  263. 9:49one way the information is coming
  264. 9:51through their delivery app system, which
  265. 9:54uses uh basically an application which
  266. 9:57uses the the email to order, and then
  267. 10:00they are collecting all the reviews in
  268. 10:02Salesforce CRM.
  269. 10:04And we need to able to sort of activate
  270. 10:07it using data from multiple of these
  271. 10:10systems. The rules that we're going to
  272. 10:12write is not just belongs to the reviews
  273. 10:15here. Like our rules here, as you can
  274. 10:17see, is two different rules here. One is
  275. 10:19if the order da- date is sort of lapsed
  276. 10:22um 60 days beyond, and also the rating
  277. 10:26or the rating is less than two. So you
  278. 10:27you can see we're creating a
  279. 10:29segmentation operation to sort of win
  280. 10:32back these customers from multiple
  281. 10:34systems. And these are extra type of
  282. 10:36problems actually uh at scale that
  283. 10:38Data3Sixty sort of solves.
  284. 10:40>> Yeah, so this is a common problem I've
  285. 10:42seen Mohit where a lot of customers may
  286. 10:45have unhappy customers they're serving.
  287. 10:48And how do you how do you figure out
  288. 10:49that's the right customer, right? When
  289. 10:50you have this data among multiple
  290. 10:52different systems, how do you know
  291. 10:54this reviewer is actual customer of you?
  292. 10:56Is that a real review? Is it a fake
  293. 10:57review? How do you figure out should you
  294. 10:59reach out to them or is it worth your
  295. 11:01time?
  296. 11:02>> Perfect, perfect. Yeah, yeah. No, this
  297. 11:03is the use case. I mean used to call
  298. 11:05this CDP use case. I I remember the term
  299. 11:08CDP use case, right?
  300. 11:11But here's here's the simple design to
  301. 11:13solve these kind of things, and I just
  302. 11:15put it out here because this is very
  303. 11:17common um you know, solution design that
  304. 11:21you um that you as solution engineers
  305. 11:24are are doing most of the time. So you
  306. 11:26will basically in Data3Sixty create data
  307. 11:29streams
  308. 11:31uh for different data um
  309. 11:34sources. In this case, it is that app,
  310. 11:37the data coming from the app,
  311. 11:39which in my case I've simplified it
  312. 11:41actually with just a CSV file. But,
  313. 11:43think about it's coming from your mobile
  314. 11:45app or something. And then
  315. 11:48the data the other data
  316. 11:51source here is the Salesforce itself
  317. 11:52that has all the reviews. So, we take
  318. 11:54all those so we create data streams, we
  319. 11:56create
  320. 11:58what we call data lake objects,
  321. 12:01and then from that we create data model
  322. 12:05objects by mapping them to the, you
  323. 12:07know, the canonical data model that
  324. 12:10Data 360 provides. So, Data 360 by
  325. 12:12default has pre-built data models for
  326. 12:16different business purposes
  327. 12:18to activate that data
  328. 12:20or unify that data. Right? So, so that
  329. 12:23is unification. So, we'll do that
  330. 12:24unification. Generally, that that is
  331. 12:26your workflow, creating streams and
  332. 12:29doing this unification or mapping it to
  333. 12:31the right canonical data model within
  334. 12:34the Data 360. And then you resolve using
  335. 12:36identity resolution. In this case, it's
  336. 12:38very simple like I've just taken email
  337. 12:40as that common thing. But, it could be,
  338. 12:43you know, multiple logic like multiple
  339. 12:44stuff. It could be very complex too.
  340. 12:47Fuzzy matches, etc. So, based on that
  341. 12:49you map them and match it and create
  342. 12:52that one single identity record
  343. 12:55from which we can create this
  344. 12:57segmentation which will be another
  345. 13:00data object called win-back diners,
  346. 13:02you know, where we sort of segment all
  347. 13:05those customers who have like sort of
  348. 13:07not ordered from Pronto app.
  349. 13:11You know, it's been like 60 days since
  350. 13:13they ordered or
  351. 13:15the rating has like sort of fallen less
  352. 13:17than two.
  353. 13:19So, this is the solution design. Now
  354. 13:22you know, because I've worked with the
  355. 13:24system, I know the solution design.
  356. 13:27but when you are solving a problem,
  357. 13:29right? And because we have different
  358. 13:31tool sets,
  359. 13:32another way to approach this kind of
  360. 13:35problem these days with the agents is
  361. 13:39ask the agent itself, like how do I
  362. 13:42solve using that MCP that we have, and
  363. 13:45the agent will definitely sort of look
  364. 13:47at all the payloads and can recommend
  365. 13:49that these are the things. Again, in
  366. 13:51this loop you still need to be in the
  367. 13:53loop because
  368. 13:54you need to judge whether whatever that
  369. 13:56agent is sort of telling you is correct
  370. 13:58or not. So, you still need to understand
  371. 14:00all of this. Uh
  372. 14:02but uh once you sort of understand, then
  373. 14:05this whole MCP uh and skills and
  374. 14:08everything becomes much more useful
  375. 14:10because then they are like, you know,
  376. 14:11power tools, which you can put them into
  377. 14:14action. I can move with a very uh
  378. 14:17you know, rapid pace. Um any thoughts
  379. 14:19here, um Anush?
  380. 14:21>> Yeah, so uh I think you really nailed
  381. 14:24that one because even for me, when I if
  382. 14:27for me, sometimes the business problem
  383. 14:29is just I want to win back these diners.
  384. 14:31Like, I do not know anything about data
  385. 14:33360. I don't know I need to unify this
  386. 14:35data. All I need to do is my goal is I
  387. 14:37need to win back these customers who
  388. 14:39have dropped off from the platform,
  389. 14:40right? And I usually pair with Claude or
  390. 14:43Codex, and I'm like, "Hey, this is my
  391. 14:45problem.
  392. 14:46This is what I have at my disposal.
  393. 14:48Help me brainstorm through the
  394. 14:49solution."
  395. 14:51And as long as you're connected to the
  396. 14:52tools to the agent, whether it's Claude
  397. 14:54or or Codex or even GitHub co-pilot,
  398. 14:58they'll be able to help you reason
  399. 14:59through your various options. And uh
  400. 15:02like Mohit mentioned, that since Mohit's
  401. 15:03a data 360 expert, he has a good grasp
  402. 15:06of what he needs to do, that is unify,
  403. 15:09resolve, and segment. But if you are new
  404. 15:11to data 360, just ask your coding agent,
  405. 15:14"Hey, this is my problem. How can I
  406. 15:16solve this?" And you'll be able to
  407. 15:17brainstorm through it.
  408. 15:19>> Yeah. Yeah.
  409. 15:20Yeah, exactly, perfect. Um so, let's
  410. 15:23let's go into that live coding segment
  411. 15:25with the coding agent
  412. 15:27and the data 360 MCP. So, we want to
  413. 15:30take this MCP server for a ride today
  414. 15:32and see how it kind of does.
  415. 15:35And again, you know, we want to be
  416. 15:37upfront honest with you. So, this is all
  417. 15:39GA, right?
  418. 15:40There are certain things in here that we
  419. 15:43are still bridging like of course data
  420. 15:45360 has a lot of like a lot of APIs and
  421. 15:48a lot of capabilities, right? And to
  422. 15:50bring all of them to the MCP, I think
  423. 15:53Anush is Anush and the team is working
  424. 15:55on. But, I would say from my experience
  425. 15:57working on this now from like last 2
  426. 15:59weeks, I can say, you know, definitely
  427. 16:01this is this is
  428. 16:03helping me like, you know, I'm able to
  429. 16:05sort of move with the speed of my intent
  430. 16:08here.
  431. 16:09You know, just knowing what I need to do
  432. 16:11and then using this MCP has has been a
  433. 16:14helper. Anush, do you have any thoughts
  434. 16:16on what are the things that you're still
  435. 16:17working on?
  436. 16:19Yeah.
  437. 16:19>> So, anything that's a UI functionality,
  438. 16:24it will be exposed to the API and you
  439. 16:26will see it in the MCP server.
  440. 16:28If there are some recent APIs, you
  441. 16:30should be seeing them shortly in the MCP
  442. 16:32server.
  443. 16:33But, there are also certain APIs or
  444. 16:35certain functionality we might not
  445. 16:37expose.
  446. 16:38An example for that is
  447. 16:40imagine you need to set up a connection
  448. 16:42which requires putting your credentials
  449. 16:44into an API or credentials into the UI
  450. 16:47to connect from system to system.
  451. 16:49>> Right.
  452. 16:49>> We might not want you to put the those
  453. 16:51credentials into the agent because then
  454. 16:54that might be stored somewhere which
  455. 16:56might lead to a security lapse.
  456. 16:59>> Yeah.
  457. 16:59>> So, you might notice very few sliver of
  458. 17:03functionality that we might
  459. 17:05intentionally keep away
  460. 17:07from the lens of
  461. 17:10security as well as governance. But,
  462. 17:12whatever other functionality that you
  463. 17:14have, you will see it in there. If we If
  464. 17:17we have missed something, please reach
  465. 17:18out to us. Let us know if something
  466. 17:20doesn't work out for you well, let us
  467. 17:21know and we'll be We'll make sure to
  468. 17:23address that as soon as possible.
  469. 17:25>> Yeah, perfect. Uh perfect. Um so first
  470. 17:28actually, so I want to show you the
  471. 17:30Pronto instance. So this is the the
  472. 17:32Pronto instance that I was talking about
  473. 17:34and uh
  474. 17:35this is basically what I'll be using for
  475. 17:37my uh building uh demos. So one of the
  476. 17:40things we will do is ask the agent to
  477. 17:42create data streams
  478. 17:44um and then create identity resolutions
  479. 17:47today with um data streams from
  480. 17:49Salesforce
  481. 17:50um
  482. 17:51pulling all the reviews. So these are in
  483. 17:52Salesforce. Now we want to pull it into
  484. 17:54data data 360 here so we can unify with
  485. 17:57the app orders data. So that's something
  486. 17:59we'll be doing.
  487. 18:00But before all this, there is some setup
  488. 18:03that is involved in um sort of setting
  489. 18:07up your org so that that MCP gets
  490. 18:09activated and you can connect to your
  491. 18:11coding agent. I want to go through that
  492. 18:14now uh with you. So uh real quick,
  493. 18:17right? Um you should find
  494. 18:20this MCP server in your API catalog. So
  495. 18:25if you go to API catalog, you will see
  496. 18:27something like data 360 server. So um
  497. 18:31again for every org, you will basically
  498. 18:35go there um and this is going to take
  499. 18:38some some time. Let me actually refresh.
  500. 18:40Sometimes it does take some time to
  501. 18:42load. But really uh what it will have is
  502. 18:44it will have a toggle. Now there is
  503. 18:47already a Salesforce CLI command too
  504. 18:51that the CLI team is exposing. It's in
  505. 18:53beta at this point through which you can
  506. 18:56um sort of not have to like go through
  507. 18:58the UI
  508. 18:59uh to activate the servers. You can just
  509. 19:02like
  510. 19:03do it through the CLI itself. So that
  511. 19:06way you can script it if you need to.
  512. 19:08Uh let's see why it's kind of hanging up
  513. 19:10here.
  514. 19:12Let's give it a shot once more.
  515. 19:15This is why I like headless actually.
  516. 19:20The headless makes it so easy. I don't
  517. 19:22have to click around and sort of wait
  518. 19:24for, you know, uh this UI to sort of
  519. 19:26load up. Uh but while that is loading
  520. 19:29up, right? Uh once that is sort of
  521. 19:32activated, which it will, uh it will
  522. 19:34show up here. But let me show you what
  523. 19:37you need to do. So so the first step is
  524. 19:38obviously to activate it. The next step
  525. 19:41here is to
  526. 19:43um create something called as external
  527. 19:46uh
  528. 19:47client credential app. And uh it's In my
  529. 19:51case, my coding agent is the
  530. 19:54Cloud Code. So for Cloud Code, I create
  531. 19:57this client credential app. Um there are
  532. 20:00certain um
  533. 20:02uh settings that you need to do, the
  534. 20:04OAuth settings, uh like the callback
  535. 20:07URL, uh the scopes, the MCP scopes that
  536. 20:11you need, and uh the the refresh token.
  537. 20:15Uh and then you will enable the the JWT
  538. 20:18token access. Um
  539. 20:21uh and uh yeah, once you do that, you
  540. 20:23will pretty much be able to like uh
  541. 20:26request for um the the keys, the
  542. 20:29consumer key. So you'll get consumer and
  543. 20:32a key and the secret. And this is very
  544. 20:34secret for for that specific org um that
  545. 20:37your clients are going to connect to. So
  546. 20:39you use that secret and then uh
  547. 20:42basically authorize this MCP server so
  548. 20:45you can connect from uh your coding
  549. 20:48agent. This is what you will going to
  550. 20:50do. Uh so this is this UI finally loaded
  551. 20:53here. So as you can see, it has three
  552. 20:55tools, the Data 360. It has three tools
  553. 20:58here. This is the server endpoint. And
  554. 21:01the three tools that we discussed is
  555. 21:02first to search that right Data 360
  556. 21:05tool, um and then get the payload
  557. 21:08example, so it can figure it out, and
  558. 21:10then it's going to execute that data 360
  559. 21:12tool.
  560. 21:13So, having said all of these, right? Um
  561. 21:17Anush actually built up a skill that we
  562. 21:20will figure it out, sort of how to like
  563. 21:22ship this to you, so that you don't have
  564. 21:24to like click around
  565. 21:26uh to set it up. It's called set up
  566. 21:28platform MCP server, basically. And this
  567. 21:31skill file I I got it from um Anush.
  568. 21:34Anush, thanks for sharing this.
  569. 21:36So, really uh Anush, you want to talk
  570. 21:38about the skill file? I think you built
  571. 21:39this, yeah.
  572. 21:40>> Uh yeah, we'll we'll find a way to share
  573. 21:42that soon. Uh the teams have been
  574. 21:44working on having this setup skill. Uh
  575. 21:47we do understand that uh setting up the
  576. 21:49external client app and connecting it to
  577. 21:52coding agent currently is a couple of
  578. 21:55steps.
  579. 21:55>> Mhm.
  580. 21:56>> And we want to solve that. So,
  581. 21:59moving forward soon, it should be as
  582. 22:01simple as you ask your coding agent set
  583. 22:04up data 360 using the skill,
  584. 22:06and it should be able to go set up the
  585. 22:08MCP server and the external client app
  586. 22:10for you.
  587. 22:11>> Yeah. Really whatever I described right
  588. 22:13now is there in the skill.
  589. 22:15>> Yeah, there's one caveat though. Make
  590. 22:17sure that you are the person setting
  591. 22:19this up as permissions to create an
  592. 22:21external client app. So, one of the
  593. 22:22things I've noticed is not everybody in
  594. 22:24your org might have that permission. So,
  595. 22:26make sure that you have access to go
  596. 22:28create the external client app.
  597. 22:29>> Yeah, perfect. And for those of you who
  598. 22:31are looking um to like sort of
  599. 22:35um
  600. 22:37understand a bit more about like
  601. 22:40how to like headlessly activate um some
  602. 22:43of these, right? Um there are
  603. 22:46uh as I said, there is one command that
  604. 22:48I can I can share with you right now.
  605. 22:51It's kind of going uh in uh I would say
  606. 22:54in in um
  607. 22:56uh in beta right now,
  608. 22:58uh kind of beta, uh but you can um sort
  609. 23:02of ask help about it. It's called SF
  610. 23:04agent MCP. Um, that's the command. And
  611. 23:08if you ask for help, it will show you
  612. 23:11um, everything. Like you can create an
  613. 23:14MCP server in that API catalog, you can
  614. 23:16delete it, you can actually fetch uh,
  615. 23:19MCP server, you can list it, you can
  616. 23:22update the server. So, this is a command
  617. 23:24that is also shipping up with the CLI.
  618. 23:26So, if you're scripting this, um,
  619. 23:28because I know a lot of lot of you
  620. 23:30actually work with scratch orgs and
  621. 23:32sandboxes, and you would want to like
  622. 23:34just automate this whole thing,
  623. 23:37um, because the moment you create a
  624. 23:38scratch org, you want these MCP servers
  625. 23:40to be activated so that you can connect,
  626. 23:43uh, easily, right? So, if you're doing
  627. 23:44that, I suggest go and look at these
  628. 23:47commands.
  629. 23:48Um, these are in beta, soon going GA,
  630. 23:50but this is going to help you in like
  631. 23:52setting up your scratch orgs or
  632. 23:54sandboxes from the CI/CD pipeline.
  633. 23:57Um, okay. So, with with having said
  634. 24:00that, let's actually move on. And so,
  635. 24:03once you do that, basically, uh, I have
  636. 24:06already set up this, um, MCP server. You
  637. 24:09can see it's already connected in my
  638. 24:10Cloud Code. I'm using Cloud Code. Um,
  639. 24:14it's one of the coding agents that I
  640. 24:15use, but again, this is MCP server. So,
  641. 24:18you can bring whatever coding agent you
  642. 24:19like, you know. Some of you I know are
  643. 24:21using Pi, some of you are using Code X,
  644. 24:24or you know, some of you are using Agent
  645. 24:26Force Wipes. We are headless, right? So,
  646. 24:28the idea with headless is this. Like
  647. 24:30this whole platform is available to any
  648. 24:32coding agent. So, but today I'm going to
  649. 24:35use the Cloud Code because that's what I
  650. 24:36I use it. So, you can see I'm connected
  651. 24:38to this Data 360 uh, MCP server. And if
  652. 24:42you actually go, you can actually view
  653. 24:44the tools. I'm already authenticated,
  654. 24:46actually.
  655. 24:47Um, so, it's all it's all well and good,
  656. 24:49but if you're not authenticated, there
  657. 24:51is one,
  658. 24:52um, command that you will need to type
  659. 24:54in into your Cloud and provide the
  660. 24:58the you know, the consumer secret and
  661. 25:00the consumer key in that command. Make
  662. 25:03sure you don't again paste the consumer
  663. 25:04key or the consumer secret into this
  664. 25:06cloud code tool. Just be on the terminal
  665. 25:09mode always when you are doing some kind
  666. 25:11of sensitive um commands, right? So, you
  667. 25:14you go on a terminal and you do it. Um
  668. 25:16but once you are connected, right? This
  669. 25:19tool now can use the MCP server. So,
  670. 25:22uh I want to show you um
  671. 25:25uh the data 360 and what I have been
  672. 25:27building with this scenario to solve the
  673. 25:29scenario so that we don't do everything
  674. 25:32here live because we'll run out of the
  675. 25:33time.
  676. 25:34So, you can see here um
  677. 25:36we have this uh diner contact stream.
  678. 25:40This stream basically is pulling data
  679. 25:42from Salesforce from contact object. And
  680. 25:45it maps to a bunch of Again, this whole
  681. 25:49thing I created um through the MCP
  682. 25:51server itself. Like I I did not
  683. 25:54go and do this 100 number of clicks. Um
  684. 25:57I asked the MCP server to create it. Um
  685. 26:00and then I said, you know, I want these
  686. 26:02these mappings for these data and it
  687. 26:04automatically figured it out like which
  688. 26:06objects is it should map to.
  689. 26:09All that Yeah.
  690. 26:11>> Mohit, before the MCP server, how would
  691. 26:13you reason through this mappings
  692. 26:15previously?
  693. 26:17Just curious.
  694. 26:18>> Yeah, so one of the things
  695. 26:21um
  696. 26:22like one of the things uh how we would
  697. 26:24reason to this is
  698. 26:27um
  699. 26:30So, there are two aspects to to
  700. 26:32reasoning here. How how um we used to do
  701. 26:35it.
  702. 26:36I think there is an auto map feature
  703. 26:38somewhere in the tool. So, once you
  704. 26:40click in the UI, it will automatically
  705. 26:42look at the data types and sort of map.
  706. 26:47But that will only work until you kind
  707. 26:49of select that data model entity. Like
  708. 26:52you will say, you know, select contact
  709. 26:54and then account contact. So, in this
  710. 26:56case, and this is a good question from
  711. 26:58business perspective, how I reason is
  712. 27:00See, in in data 360, there are only
  713. 27:03two key things. Whether you want to do
  714. 27:06the unification and this identity
  715. 27:09resolution
  716. 27:10by individual, or you want to do it
  717. 27:14by the entity, like account, right? And
  718. 27:18or by business, right? Um so, usually it
  719. 27:21is mapping to account contact and
  720. 27:23contact point addresses and point phone
  721. 27:26and individuals, but in matches by data
  722. 27:29types. And of course, I have to go by
  723. 27:31hand and map a lot of these, yeah.
  724. 27:34>> Yeah, you know, I noticed that at at
  725. 27:36least for me personally, mapping was a
  726. 27:38tricky flow because a lot of this had to
  727. 27:41be hand mapped.
  728. 27:42>> Mhm.
  729. 27:42>> Uh so, how was your experience from
  730. 27:45how you did this previously and now with
  731. 27:47MCP server?
  732. 27:49>> Yeah, I would say it was good. All I did
  733. 27:52was um just ask the the cloud code. I'm
  734. 27:55going to show one of them. So, this this
  735. 27:57I built it already for the contact
  736. 27:59streams and we have one for the app
  737. 28:00order stream. So, let's do one one from
  738. 28:03the tool, so everybody can actually see.
  739. 28:06So, one of the other things as I said,
  740. 28:07we want to pull is all that review. So,
  741. 28:09there is a review object. So, typically
  742. 28:11how you would build is, you know, you
  743. 28:13will go to new and you know, you will do
  744. 28:15so many clicks. You will say yes, CRM
  745. 28:16objects and wait for this screen to load
  746. 28:20load and then you will be like view the
  747. 28:22objects. There are too many clicks, you
  748. 28:24know, too many clicks. My org is I would
  749. 28:26say reasonable, but still you will you
  750. 28:28will have to search, etc. etc. find a
  751. 28:30lot of these things, right? Uh a lot of
  752. 28:32clicks. Um
  753. 28:34So, what I'm going to do is I actually
  754. 28:36wrote a a spec of what I want. Um let's
  755. 28:40let's look at that spec of the changes I
  756. 28:42want. So, one of things I'm doing now is
  757. 28:45this review uh data stream, um
  758. 28:48basically. So, again, uh
  759. 28:50again I'm asking what I need to do here
  760. 28:52in this in this case, right? So, the
  761. 28:55whole idea here with this spec and you
  762. 28:58know, all the features that got added in
  763. 29:00VS code was being able to view this
  764. 29:02markdown now.
  765. 29:03So, really what I'm saying is look,
  766. 29:04contact and app order are ready. I need
  767. 29:07the review signals ingested into new
  768. 29:09data stream. And then this is how I like
  769. 29:12to write my specs actually. Just just
  770. 29:14So, it is consistent. So, you know,
  771. 29:16Anush, you it need not be like this. You
  772. 29:18can just write it however you want. It's
  773. 29:21It's up to you. But I think it makes it
  774. 29:23much easier for me also to think, right?
  775. 29:25In this case I'm saying, you know, once
  776. 29:27the system will create a data stream for
  777. 29:29the CRM object
  778. 29:31and lands as a DLO carrying the rating,
  779. 29:33review, order, and contact. And then
  780. 29:36this is the only injection artifact this
  781. 29:38change creates. And this is built using
  782. 29:41that MCP discovery loop, right? Not
  783. 29:44giving again, not giving the payload
  784. 29:45etc. but just saying it. Scenario is,
  785. 29:48you know, review DLO is populated and
  786. 29:50carries the signal. So, this is one
  787. 29:51scenario. The other scenario is even
  788. 29:54data stream is provisioned correctly.
  789. 29:55Like once it's done, make sure it's kind
  790. 29:57of, you know, provisioned correctly.
  791. 29:59So, this is basically
  792. 30:01it. And you know, if you know certain
  793. 30:03things, like I know there are these many
  794. 30:06DLOs, these many rows, so I can also
  795. 30:08say, right? When you query these many
  796. 30:10rows are there.
  797. 30:12If you know,
  798. 30:13sometimes whatever you know, right? That
  799. 30:16becomes a deterministic signal here.
  800. 30:18Um
  801. 30:19So, yeah, simple simple requirement, you
  802. 30:22know, that I want. Now I'm going to just
  803. 30:25ask that use this review data stream
  804. 30:27spec, right? And create a data stream.
  805. 30:30So, we'll we'll we'll we'll chat about
  806. 30:32it, but let's submit this work to the
  807. 30:33agent here. So, we're going to say
  808. 30:37create
  809. 30:39a data stream for the reviews.
  810. 30:43The specs are here.
  811. 30:47Um yeah, so I'll uh
  812. 30:51I'll actually submit it right in the
  813. 30:53auto mode
  814. 30:54and let the agent kind of figures it out
  815. 30:58and create that uh stream.
  816. 31:00Um
  817. 31:01Yeah, so so Anush, you're you're asking
  818. 31:03me something. Uh it kind of works
  819. 31:05through.
  820. 31:05>> Yeah, I I was cu- I was curious because
  821. 31:07like I wanted to try to gauge what your
  822. 31:11workflow was before versus what it is
  823. 31:13after. Like how have you changed the way
  824. 31:16in which you operate, right? Like how
  825. 31:17much of your previous context was
  826. 31:19important and what can users learn from
  827. 31:22the way you have changed your uh usage
  828. 31:24of Data 360?
  829. 31:26>> Yeah, yeah. So the biggest the biggest
  830. 31:28unlock for me doing this is see, I'm no
  831. 31:30more a blocker. You know, we have to
  832. 31:33think in these terms, right? If I sit on
  833. 31:35the computer and then like just click
  834. 31:37around like 50 number of clicks,
  835. 31:40it's
  836. 31:42like I I'm I'm the blocker now to to
  837. 31:44execute that job, right? And if I know
  838. 31:47that the agent can like, you know,
  839. 31:49reliably go and do certain aspects of
  840. 31:51this, right? Uh I'll probably kick that
  841. 31:54load and then move on, you know, like
  842. 31:56sort of do something uh else, right? Um
  843. 31:59so that that has been operational change
  844. 32:02uh from me is like now I'm spending more
  845. 32:04time instead of like, you know, clicking
  846. 32:07around, I'm more spending more time
  847. 32:09doing the the the design thinking
  848. 32:13um mostly around, okay, how do I solve
  849. 32:15this business problem? What are the
  850. 32:16tools I need? And writing it um in um
  851. 32:19natural language so I can then queue
  852. 32:22that work to the agent. And then once
  853. 32:24the agent kind of does, you know, my
  854. 32:26work is more of verification. So as you
  855. 32:28can see, my work is sort of moved moved
  856. 32:31from like
  857. 32:32implementing, right? I have more time to
  858. 32:35do the other side of the work now. Like
  859. 32:36the the planning, you know, I spend more
  860. 32:38time doing that planning. After the work
  861. 32:40I spend more time doing the doing, you
  862. 32:43know, judgment and all these kind of
  863. 32:44capabilities. So So that is what I've
  864. 32:47been doing. Yeah.
  865. 32:48>> So you moved like you moved up the
  866. 32:49stack. You have expanded your scope from
  867. 32:51being a just a developer to being a
  868. 32:53developer but also like a
  869. 32:55uh
  870. 32:55directly responsible individual for
  871. 32:58coming together with a business plan and
  872. 33:00evaluating if it is successful or not.
  873. 33:02>> Absolutely. Absolutely. And you know,
  874. 33:05this is one terminal that is running as
  875. 33:07you can see. I My my day now is if if
  876. 33:10somebody asked me to do something,
  877. 33:12right? Usually I have like five or six
  878. 33:13terminals running and I'm just running
  879. 33:15you know, sort of managing those
  880. 33:16terminals or those agents in this case
  881. 33:19we can say, right? Um so you see So so
  882. 33:22Anush, let's talk about you know, what
  883. 33:24happened here, right? So you see what
  884. 33:25the agent did is it said it has the few
  885. 33:28picture. It knew that the object name is
  886. 33:31review __c engagement stream. It began
  887. 33:35that discovery loop, which is basically
  888. 33:37searching and pulling down the exact
  889. 33:39schema. So now it has the schema.
  890. 33:42So now it knows that there is a home
  891. 33:44connection already in the org. That is
  892. 33:46so awesome, right? It knows that there
  893. 33:48is that CRM connection is there and if
  894. 33:52it would have the connection would have
  895. 33:53been missing, it will tell me that go
  896. 33:55first do that connection. Um so I can go
  897. 33:58go do the connection. So I will be like
  898. 33:59sort of blocked at this point, right?
  899. 34:01But now since it has a connection, it's
  900. 34:02able to go move. So it's now able to go
  901. 34:05and look at all the things. It looked
  902. 34:07existing stuff that are there in the in
  903. 34:09the org. Um it looked into all the
  904. 34:12prerequisite. It confirmed the the the
  905. 34:14field and now it's calling the data 360
  906. 34:18um APIs to I think hopefully this is the
  907. 34:21next step is to create that um
  908. 34:23that uh data streams. In fact, I think
  909. 34:25it's creating. Let's Let's look at see
  910. 34:28if So you see?
  911. 34:29There you go. You see it actually
  912. 34:31created it. So now it's going to
  913. 34:33schedule the I think it's going to the
  914. 34:35next thing is it's going to schedule the
  915. 34:36run, which is to like, you know, get all
  916. 34:39the
  917. 34:40the data
  918. 34:41and schedule that that run.
  919. 34:44So, it can get all the data into Data
  920. 34:46360. And
  921. 34:48it did it so, like, you know, magically,
  922. 34:50but
  923. 34:52this this would have been some clicks
  924. 34:54actually, you know. So, it says, you
  925. 34:55see, run triggered successfully. Let me
  926. 34:57poll the DLO account until it populates,
  927. 35:00right?
  928. 35:00So, I'm pretty sure it will now go and
  929. 35:03see, it moved the status to pending. It
  930. 35:06means it knows that, okay, next I need
  931. 35:08to kick this job. So, we should see some
  932. 35:10refresh history. It's not here yet, but
  933. 35:13it's the next step. So, once it does, I
  934. 35:16think the next step I'm curious to see
  935. 35:18is it does the data mapping by itself.
  936. 35:20If it's not, maybe that's one more
  937. 35:21prompt. And from there I will have the
  938. 35:24review data already synced up, sort of
  939. 35:27without me having to like click, just by
  940. 35:29intent. I'm I'm moving. Yeah.
  941. 35:31>> Um meanwhile, I'm also curious,
  942. 35:34Moeit, uh
  943. 35:35how did you which model did you pick for
  944. 35:37this and why did you choose the specific
  945. 35:39model? Do you have an insight on your
  946. 35:41usage and what you found works well for
  947. 35:43you?
  948. 35:44>> Yeah, yeah, exactly. So, this one the
  949. 35:47the model that I have picked for this is
  950. 35:51uh Opus 4.8.
  951. 35:53Um Opus 4.8
  952. 35:56is
  953. 35:57decent model. And the the best thing
  954. 35:59about Opus 4.8 to me is
  955. 36:02it somehow magically, like, knows a lot
  956. 36:05of Data 360 too. You know, I played with
  957. 36:07this model a lot. The there is that
  958. 36:09reasoning traces in there, you know,
  959. 36:11that that understands it does understand
  960. 36:14the terminology of Data 360 to some
  961. 36:16extent, right? And again, uh you know,
  962. 36:19your business might be different and
  963. 36:21that is the the ontology layer, you
  964. 36:23know, sometimes you will have to come
  965. 36:25and write some ontology layer on top of
  966. 36:29um your
  967. 36:30you know your work that you already have
  968. 36:32like lot of you would have already built
  969. 36:33a lot of these things right if I'm not
  970. 36:35wrong you you'd have built data streams
  971. 36:37and data schemas and your you have your
  972. 36:40own business needs. So as long as you
  973. 36:42can express that in markdown files or
  974. 36:46someway to the agent and sort of ground
  975. 36:48that the agent will like have the
  976. 36:50context at that point and because it
  977. 36:53understands the technology of data 360
  978. 36:55to a lot of extent
  979. 36:57it will be able to like sort of make or
  980. 36:59reason through the task really well with
  981. 37:01these MCP servers of course helping
  982. 37:03right. One question I have for you
  983. 37:07Anush is
  984. 37:08this MCP server right um
  985. 37:11is it calling the APIs that our
  986. 37:13developers can also look at somewhere?
  987. 37:16Is it calling the same APIs?
  988. 37:17>> Yeah
  989. 37:19behind behind the scenes the MCP server
  990. 37:21does call a middleware that does call
  991. 37:23the same APIs that are exposed to you
  992. 37:26which is why the
  993. 37:27>> APIs
  994. 37:29>> Yeah.
  995. 37:30>> Okay so so the connect rest APIs right
  996. 37:32so yeah if you are a developer again we
  997. 37:34have the developer docs here and you can
  998. 37:37always go there and look at the connect
  999. 37:39APIs
  1000. 37:41the ones I'm sure it's actually um
  1001. 37:44hitting right now is I think that the
  1002. 37:46data stream APIs right
  1003. 37:50that is what it is hitting right now
  1004. 37:52but again you see the orchestration of
  1005. 37:54these APIs what needs to happen when um
  1006. 37:59I mean of course for repeatable use
  1007. 38:00cases I would also like if your use case
  1008. 38:03is too complex you know there's no harm
  1009. 38:06in like um you know sort of
  1010. 38:08deterministically writing some of these
  1011. 38:10code or asking an agent to write it and
  1012. 38:12then you run it via the agent. All those
  1013. 38:14kind of workflows are definitely
  1014. 38:16possible but the key point here I want
  1015. 38:18to make out is the agent like sort of
  1016. 38:22loves orchestrating these things. I
  1017. 38:23mean, it has with MCP like those three
  1018. 38:26tools that this team has delivered,
  1019. 38:28right? It knows now these APIs really
  1020. 38:32well and it can reason through that. And
  1021. 38:33the you know, if that makes sense,
  1022. 38:35Anush.
  1023. 38:37Let's see if there are any questions
  1024. 38:38that are coming. I haven't got a chance
  1025. 38:40to look at the the chat here.
  1026. 38:42>> Yeah, that's a good idea.
  1027. 38:44>> Yeah, let's look at
  1028. 38:46um
  1029. 38:47um
  1030. 38:48chat here.
  1031. 38:50Uh is there a way that we can rate limit
  1032. 38:53the calls to these MCP servers?
  1033. 38:56Um Uh so, why is that? So, somebody is
  1034. 38:59asking Himanshu. Um let's see. I'll I'll
  1035. 39:02put the question here so we can all all
  1036. 39:04see here. So, Himanshu is saying is
  1037. 39:07there a way we can rate limit the calls
  1038. 39:08to MCP server APIs in case MCP server
  1039. 39:10connected to the autobots which are used
  1040. 39:12by internal business user and not just
  1041. 39:14coding agent. Thoughts, Anush, on this?
  1042. 39:17>> I believe
  1043. 39:19there is no way to explicitly rate limit
  1044. 39:21the MCP server today.
  1045. 39:23>> Uh-huh.
  1046. 39:24>> On the
  1047. 39:25Salesforce side.
  1048. 39:27>> Mhm.
  1049. 39:27>> But that is we are building governance
  1050. 39:29functionality into that as we go along.
  1051. 39:32But on the client side, while you
  1052. 39:34connect an MCP server to a specific
  1053. 39:36agent, you can always put limits on the
  1054. 39:38client side to figure out, "Hey, do not
  1055. 39:40make more than these number of calls,
  1056. 39:42right?" So, there are client level
  1057. 39:44guardrails. Those are not great yet, I
  1058. 39:46think. But we will be working on having
  1059. 39:49better governance and control for
  1060. 39:51Salesforce admins to have a better
  1061. 39:53organization of how much of an MCP
  1062. 39:56server do you want the users to use.
  1063. 39:58>> Mhm. Mhm. And that also will come with
  1064. 40:01more uh maybe usage metrics and traces
  1065. 40:03and things like that. So, we're still
  1066. 40:04working on some of those governance
  1067. 40:06features.
  1068. 40:07>> Perfect. Perfect.
  1069. 40:09Uh somebody is asking this question,
  1070. 40:11which is really interesting. How does
  1071. 40:13this differ from agent for Salesforce
  1072. 40:15worker?
  1073. 40:16>> So, that's a great question.
  1074. 40:19The MCP server will work with anything
  1075. 40:22that is an MCP client.
  1076. 40:24Now, whether that is Cloud, Codex,
  1077. 40:26GitHub Copilot, whether it's an MCP
  1078. 40:29client, this will work with it,
  1079. 40:30including Agent for Swipes.
  1080. 40:33>> Mhm.
  1081. 40:34>> Agent for Swipes co-worker today is not
  1082. 40:36yet an MCP client.
  1083. 40:38>> I see.
  1084. 40:39>> So, today you will not be able to
  1085. 40:40connect it. But, that doesn't mean in
  1086. 40:42the future you won't be able to. In the
  1087. 40:44future, this could be part of Agent for
  1088. 40:46Swipes co-worker, whether implicitly
  1089. 40:48from our integration or explicitly from
  1090. 40:51your integration.
  1091. 40:52>> Mhm. So,
  1092. 40:53>> uh we're not there yet.
  1093. 40:55Uh the other thing is like
  1094. 40:57you can build your own agent. You don't
  1095. 41:00have to go through any coding agent. If
  1096. 41:01you have your own agent within Pi open
  1097. 41:03code, any MCP client will work with the
  1098. 41:06Data 360 MCP server.
  1099. 41:09>> Mhm. [clears throat] Mhm. Perfect.
  1100. 41:10Perfect. Now, we know this loads takes
  1101. 41:12some time, especially for the first time
  1102. 41:14because it has to bring in all the all
  1103. 41:16the data. Let Let it Let it kind of run,
  1104. 41:19and uh we can um sort of keep moving.
  1105. 41:21So, so really, you know, for this
  1106. 41:23requirement though, my only thing will
  1107. 41:25be build this data stream, right? And
  1108. 41:27then sort of map that to the
  1109. 41:30uh engagement object engagement um
  1110. 41:33type of object category. The other
  1111. 41:35interesting thing to me was when this
  1112. 41:39like when when this data stream got
  1113. 41:40created, it immediately picked up the
  1114. 41:44the category as engagement. Which was
  1115. 41:46interesting. Like, it knew from the
  1116. 41:48business requirement that, you know, the
  1117. 41:50category will not be profile. And this
  1118. 41:52is one of the fundamental decisions that
  1119. 41:54we do in in Data 360 because everything
  1120. 41:57including how we can we are charged,
  1121. 42:00etc. depends on these kind of aspects,
  1122. 42:02like how how are you categorizing the
  1123. 42:04data, right? Um so, it did pick the
  1124. 42:06right stuff, and it's keep on moving. Of
  1125. 42:09course, this is the first it's a data
  1126. 42:10load, so it's going to take some time um
  1127. 42:12behind the scenes running those bad
  1128. 42:14jobs. Oh, we have the data. You see? Oh,
  1129. 42:16no, the fields. Sorry, is the field.
  1130. 42:19Okay, it's going to it's going to take
  1131. 42:20some time,
  1132. 42:21but you can see it's it's active.
  1133. 42:25And it's going to run. The data stream
  1134. 42:27is going to run. And it created the data
  1135. 42:29lake object. So, as I was clicking
  1136. 42:31through, you know, so you see there is a
  1137. 42:32data lake object that got created. So,
  1138. 42:35next would be, you know, DMO creation.
  1139. 42:37And you know, it itself puts like this.
  1140. 42:41I think this is a feature of the harness
  1141. 42:42itself within cloud code. It can like,
  1142. 42:45you know, put itself into the sleep
  1143. 42:46mode. So, it just sleeps and waits
  1144. 42:49for the task completion.
  1145. 42:51So, that's that's perfect. It's it's
  1146. 42:54working it's it's way out. Um
  1147. 42:57Okay, so the other thing that we wanted
  1148. 42:59to sort of discuss here with this
  1149. 43:04is
  1150. 43:05this whole idea of skills, right? Talk
  1151. 43:08us more through that, Anush, like the
  1152. 43:11skills, right? I I see developers more
  1153. 43:13often getting confused like what is
  1154. 43:15skills, what is MCP, do we need MCP
  1155. 43:20at all when we use skills? Um
  1156. 43:23etc. Like, let's let's get to the bottom
  1157. 43:25of that.
  1158. 43:26>> So,
  1159. 43:28I think in the industry there is still
  1160. 43:30some flux going on right now.
  1161. 43:32>> Mhm.
  1162. 43:32>> Um the way we see it is I think there
  1163. 43:35are two differing standards right now
  1164. 43:37coming. You could say one is uh the
  1165. 43:40skills
  1166. 43:42and one is your agent plugins. You might
  1167. 43:44have seen this, right? I think
  1168. 43:46ChatGPT or code it started with agent
  1169. 43:48plugins. A plugin can contain MCP and
  1170. 43:51skills.
  1171. 43:52And then on the other end you have MCPs,
  1172. 43:54which in the next spec will have skills
  1173. 43:56themselves.
  1174. 43:57So, we are still observing that. We will
  1175. 43:59support whichever is the most convenient
  1176. 44:02for our users.
  1177. 44:03>> Mhm.
  1178. 44:04>> So, that is the direction we're seeing
  1179. 44:05things go.
  1180. 44:07>> Okay.
  1181. 44:08>> when we say skills,
  1182. 44:10skills are a way to provide just-in-time
  1183. 44:13context
  1184. 44:14>> Mhm.
  1185. 44:15>> to the agent of your choice.
  1186. 44:17>> I see. I see.
  1187. 44:18>> What do I mean by just-in-time context?
  1188. 44:21That means whenever you have a
  1189. 44:23successful workflow where you do not
  1190. 44:25want the agent to freestyle reason
  1191. 44:28through a problem,
  1192. 44:29>> Mhm.
  1193. 44:30>> but you want it to follow a specific
  1194. 44:32workflow that you know works well,
  1195. 44:34>> Mhm.
  1196. 44:35>> you ask it to invoke a certain skill.
  1197. 44:38>> Got it.
  1198. 44:39>> Now,
  1199. 44:40for Data 360, it could be something like
  1200. 44:44we know best how to create certain
  1201. 44:46segments.
  1202. 44:47>> Mhm.
  1203. 44:48>> We know how to do transforms.
  1204. 44:50>> Right.
  1205. 44:50>> If you ask an agent to do transforms, an
  1206. 44:53agent might reason through all the world
  1207. 44:55knowledge it has
  1208. 44:56>> Mhm.
  1209. 44:57>> that is relevant and irrelevant to Data
  1210. 44:59360.
  1211. 45:00>> Got it.
  1212. 45:01>> And then it will consume tokens. It will
  1213. 45:04take a lot of tokens.
  1214. 45:06It might still get it right,
  1215. 45:08but then you're spending a lot of tokens
  1216. 45:09doing it.
  1217. 45:10>> Right. Right.
  1218. 45:11>> And it becomes slightly more
  1219. 45:13non-deterministic.
  1220. 45:14>> Mhm.
  1221. 45:15>> So
  1222. 45:16Yeah. So when the
  1223. 45:20expert in a certain product, now that
  1224. 45:22could be the Salesforce product
  1225. 45:24developers or engineers
  1226. 45:26>> Mhm.
  1227. 45:26>> or the Salesforce implementers in the
  1228. 45:28community, when you know how best to
  1229. 45:31implement a certain thing and you do not
  1230. 45:33want the agent to do some freestyle
  1231. 45:35reasoning,
  1232. 45:37codifying your workflow in the form of a
  1233. 45:39skill and then invoking that will
  1234. 45:41provide great results.
  1235. 45:43>> I see. Okay.
  1236. 45:45>> One of the trend I've seen is also that
  1237. 45:48if you pick a really good model with
  1238. 45:50high reasoning, it can actually do most
  1239. 45:53things without skills.
  1240. 45:54>> Right.
  1241. 45:55>> But then that becomes expensive,
  1242. 45:56consumes tokens.
  1243. 45:58Now the question for me is how do you
  1244. 46:00take like a Claude Sonnet or Haiku model
  1245. 46:04that is really low cost and make it
  1246. 46:06perform at the same level of Opus.
  1247. 46:08>> Right.
  1248. 46:09>> So,
  1249. 46:10cost is one place where skills are
  1250. 46:11helpful, speed is another place where
  1251. 46:13skills are helpful, and having more
  1252. 46:15deterministic execution is where skills
  1253. 46:17are helpful.
  1254. 46:18>> Got it.
  1255. 46:19>> Got it.
  1256. 46:19>> Yeah, that was enlightening. That was
  1257. 46:21enlightening. Now, while you are there,
  1258. 46:23you know, one of the things is you see
  1259. 46:25it actually pulled the record. So, that
  1260. 46:27job was success.
  1261. 46:29So, it it has a refresh history now. So,
  1262. 46:32that job is successful now. So, it's
  1263. 46:35moving on to the next thing. So, now it
  1264. 46:37says the this is all thing, you know,
  1265. 46:39these are scheduled synchronously. It's
  1266. 46:42able to infer the conclusion also,
  1267. 46:44right? Um
  1268. 46:47So, uh batch is still pending. Now,
  1269. 46:50at this time, you know, because it is
  1270. 46:52like in the sleep mode, you see, I can
  1271. 46:54see in the UI that it's all successful.
  1272. 46:56But, because the agent sort of put
  1273. 46:57everything in into the sleep mode, um
  1274. 47:00it's it's going to sleep over for like
  1275. 47:02these these many time. And then it's
  1276. 47:04going to go back and query and uh
  1277. 47:07uh sort of tell that, "Okay, this is
  1278. 47:09this is successful. Let's let's move on,
  1279. 47:11right?"
  1280. 47:12Um what I'm going to do is
  1281. 47:14of course, stop it because we know it's
  1282. 47:17successful. I'm going to say it's
  1283. 47:19successful right now.
  1284. 47:22Um
  1285. 47:23and uh can you move ahead?
  1286. 47:27Mapping, creating DMOs, and mapping it,
  1287. 47:31right?
  1288. 47:32Um
  1289. 47:33Now, again, this is because I know the
  1290. 47:36next step, so I'm able to tell. But, if
  1291. 47:38you leave it, it is automatically going
  1292. 47:40to take that step, you know, because
  1293. 47:42again, it has the reasoning like this
  1294. 47:45this Claude knows that after this, this
  1295. 47:47is the step that needs to happen, right?
  1296. 47:49Um
  1297. 47:50so, again, it's going to go and work
  1298. 47:52this out. It should
  1299. 47:54Now, it should be very simple because
  1300. 47:55now the next step it's going to do is,
  1301. 47:57you know of course map it right like if
  1302. 47:59I if I do I will need to
  1303. 48:01pull that DMO but I don't think I have
  1304. 48:03the DMO here.
  1305. 48:05Let's see.
  1306. 48:07You see it's going to every click takes
  1307. 48:09some time right in data 360 because it's
  1308. 48:12it has to search through all the objects
  1309. 48:13is a big system.
  1310. 48:15Um so
  1311. 48:17I don't think I have the diner review
  1312. 48:19object it.
  1313. 48:22You see I don't have the review object
  1314. 48:24okay we don't have the review object it.
  1315. 48:25So the agent is going to create it. So
  1316. 48:28now it's calling you see it's call it's
  1317. 48:29called four times.
  1318. 48:31Uh let me confirm the review count then
  1319. 48:33move on right up in a knot to build a
  1320. 48:35unified DMO it's to use this mapping on
  1321. 48:37the contact to auto provision and
  1322. 48:39standard and then map to the other
  1323. 48:40sources.
  1324. 48:41Now it's going to it's going to do. So
  1325. 48:43this is a
  1326. 48:45I think it learned from somewhere the
  1327. 48:46reasoning here for harmonization. Um
  1328. 48:49that it always needs to look at what is
  1329. 48:51there. So it's the the first reasoning
  1330. 48:53it did is like yeah I need to be able to
  1331. 48:55map using the standard mappings.
  1332. 48:58But I'm pretty sure it will try to find
  1333. 49:00out and then it says no there's no way
  1334. 49:02to map this review object to contact
  1335. 49:04because it's not same. It's going to
  1336. 49:06reason like that. Um it's very different
  1337. 49:08than how I think actually
  1338. 49:11uh but you can see that it actually
  1339. 49:13comes to the same conclusion maybe doing
  1340. 49:15three or four more turns on it.
  1341. 49:18Um
  1342. 49:19but it's it's it's it's different. So
  1343. 49:22you can see it will now create the
  1344. 49:24mappings. It will take some time I'm not
  1345. 49:25going to do it manually. The the agent
  1346. 49:27is going to create the mapping. So once
  1347. 49:29it's create the mapping I'll ask it to
  1348. 49:31create the
  1349. 49:32the identity resolution. Um again uh
  1350. 49:35just for all of you right I will want to
  1351. 49:37show really like what what would be my
  1352. 49:40next step here or what is it that we're
  1353. 49:42going to do here. So you can see the
  1354. 49:44review you we have the reviews object we
  1355. 49:46have the contact team and we have the
  1356. 49:47app order object right so we have all of
  1357. 49:49these. So the next step is to actually
  1358. 49:52use all of this and sort of map it to
  1359. 49:55individual objects within
  1360. 49:57um
  1361. 49:58within Salesforce and sort of relate all
  1362. 50:00of these
  1363. 50:01um
  1364. 50:02and then uh from there, the idea is to
  1365. 50:05create this uh identity resolution job,
  1366. 50:07this one.
  1367. 50:09Now, I just need to wait uh for um the
  1368. 50:12DMOs to be created here. So, standard
  1369. 50:14mapping succeeded. Now, I need to map
  1370. 50:17the app orders DLO. So, now it's
  1371. 50:19mapping. So, you see the mappings is
  1372. 50:20happening now. So, once it maps, my next
  1373. 50:23step would be again, as I said, create
  1374. 50:25the
  1375. 50:26So, you see the SSO individual is
  1376. 50:27populated. Now, let's get the app order
  1377. 50:30DLO.
  1378. 50:31And yeah, next my next intent will be to
  1379. 50:33create that identity resolution job and
  1380. 50:35it's going to go go create it. Now, I
  1381. 50:38know a lot of you would have done
  1382. 50:39identity resolution and you know the
  1383. 50:41pain. It's like you have to click at
  1384. 50:44least 10 to 12 or maybe more clicks in
  1385. 50:47here to actually do certain things in
  1386. 50:50Data 360. And I'm so glad that this MCP
  1387. 50:53server is here and it's able to like
  1388. 50:56you know, sort of automate all of this
  1389. 50:58and uh instead of me sitting through and
  1390. 51:00clicking
  1391. 51:01um the system can like reason through.
  1392. 51:04Uh let's see if there are any any
  1393. 51:06questions that are coming.
  1394. 51:09Uh
  1395. 51:11using headless 360, how to pass data
  1396. 51:13effectively react as a front-end
  1397. 51:16framework?
  1398. 51:17Uh very simple, actually. Um this is
  1399. 51:20something I've I've myself done it. It's
  1400. 51:23everything in Data 360 can be queryable
  1401. 51:26through Apex. So, Apex recently, and let
  1402. 51:29me show you guys this one.
  1403. 51:31Um Apex connect query.
  1404. 51:34Um
  1405. 51:36So, Apex um
  1406. 51:38uh connect query for Data 360. Apex
  1407. 51:41actually launched uh a new modern
  1408. 51:45um query class, actually, that you can
  1409. 51:48use it to just query yeah, it's called
  1410. 51:51SFQL query namespace. This is a new
  1411. 51:54namespace actually. Um, I've used it
  1412. 51:57often. So, once you know your tables,
  1413. 52:00you're just writing SQL statements
  1414. 52:01within Apex.
  1415. 52:03Basically very simple. Try it out and
  1416. 52:05then you know, you can build react apps
  1417. 52:07or whatever on Salesforce. And then just
  1418. 52:09use these to query um, from data 360. We
  1419. 52:12we made it much more easier now to just
  1420. 52:15natively sort of use it. Uh, let's see
  1421. 52:18if there are any any other questions
  1422. 52:19here.
  1423. 52:20Um, this is one. How is your diner data
  1424. 52:24stream able to function when um, fields
  1425. 52:28have not been mapped? You might be in
  1426. 52:30the process of configuring it.
  1427. 52:32Um, yeah. So, the diner data stream So,
  1428. 52:35first you create the data stream and of
  1429. 52:37course it's not mapped, but
  1430. 52:39um, you know, yeah, we need the mapping
  1431. 52:41before we do the identity resolution.
  1432. 52:42That's why I'm waiting for it. Let's see
  1433. 52:44if it is actually done.
  1434. 52:47Uh, let's see where the agent is.
  1435. 52:50Um, yeah, agent is working through it.
  1436. 52:52You can see
  1437. 52:54uh,
  1438. 52:56most of the effort creating. So, it is
  1439. 52:59So, the other thing that Anusha is saw
  1440. 53:01this is one of the wonderful things is
  1441. 53:03actually goes back and looks what is
  1442. 53:05there in the org.
  1443. 53:06>> Yeah.
  1444. 53:07>> Um,
  1445. 53:08>> That itself was previously taking me a
  1446. 53:11lot of time, right? Like whenever I need
  1447. 53:12to go to context switch between what I'm
  1448. 53:14doing and go back to the org and go back
  1449. 53:17to my documentation, it's saving me a
  1450. 53:19lot of that time recently.
  1451. 53:21>> Yeah, and I think why it is doing is
  1452. 53:23because I have this skill called reuse
  1453. 53:25before reinvent. So, this is one of the
  1454. 53:27skills I wrote it in my harness. And my
  1455. 53:31plan is very simple is like anytime the
  1456. 53:33agent builds something, I wanted to go
  1457. 53:36and see if there is anything it can
  1458. 53:38reuse.
  1459. 53:39It can backfire also sometimes. Like it
  1460. 53:42will try to It will try to use um things
  1461. 53:46that it shouldn't. Um but I have the
  1462. 53:49skill. So, So, again, uh this is very
  1463. 53:51interesting thing, right, with the the
  1464. 53:53skills concept here.
  1465. 53:54>> Yeah.
  1466. 53:55>> Like, yeah, it all depends on you. Like,
  1467. 53:57the hardness
  1468. 53:58I mean, there is a thing like there will
  1469. 54:00be a a baseline hardness that somebody
  1470. 54:03can provide. But, then how you work um
  1471. 54:07if you want to like
  1472. 54:09like keep that non-shareable, then you
  1473. 54:12can do that as well. Like, you can
  1474. 54:14create your own skills and then use it
  1475. 54:16within your hardness. So, that's that's
  1476. 54:18very powerful
  1477. 54:19um capability. So, yeah, it's calling
  1478. 54:23It's calling It's It's doing some some
  1479. 54:25stuff here. Let's see.
  1480. 54:28Yeah, it's it's doing It's basically
  1481. 54:30going to create that object and then I
  1482. 54:32think, yeah, the identity resolution, I
  1483. 54:34think it will
  1484. 54:36it will create. I I need to ask it to
  1485. 54:38create identity resolution. Currently,
  1486. 54:40it is actually listing.
  1487. 54:41So, it's asking me. See, the standard
  1488. 54:44DMOs. So, it kind of asking me the
  1489. 54:46standard DMOs and the contact order
  1490. 54:48mapping already exist. What should I
  1491. 54:50build next? So, create that DLO. Yeah.
  1492. 54:55Yeah. Skip further mapping and build the
  1493. 54:58IR. Of course, we can't build the IR
  1494. 55:00without we know that we have to map the
  1495. 55:02the DLOs onto the individual with the
  1496. 55:04party and the rating. Uh this is the
  1497. 55:07missing DMO. Yeah. So, I'm going to go
  1498. 55:09with that option.
  1499. 55:10This is that loop. This is what I call
  1500. 55:13it loop. If you are upfront like
  1501. 55:17doing all the decisioning
  1502. 55:19you will never get in caught into these
  1503. 55:21kind of loops, right? We can get rid of
  1504. 55:23these kind of loops where it ask your
  1505. 55:24feedback, but sometimes you can design
  1506. 55:26loops where, you know, you are also part
  1507. 55:29of the loop. So, yeah. So, now it now
  1508. 55:32it's going to build that DMO. I think
  1509. 55:34the next step is that. And if I leave my
  1510. 55:37agent, it will next build identity
  1511. 55:39resolution, segmentation, whatnot. In
  1512. 55:41fact, you know, I had a spec that I I
  1513. 55:44ran through this. Uh let me show you
  1514. 55:46that spec.
  1515. 55:48It's called Diner Winback. And uh just
  1516. 55:51because, you know, this whole thing can
  1517. 55:53take like hours of long-running task, we
  1518. 55:56we are short of time here. But, this is
  1519. 55:59the spec I wanted to show that I
  1520. 56:01actually ran in this org. So, I
  1521. 56:04indicated all the source system. I
  1522. 56:06exactly told what I need to do.
  1523. 56:09Um I told it how to segment it, um how
  1524. 56:12to unify it, uh and then my scope of the
  1525. 56:15work and the test boundaries, and my
  1526. 56:17acceptance criteria and my assumptions.
  1527. 56:19And uh you know, some things like, you
  1528. 56:22know, I wanted to
  1529. 56:24like keep a a companion doc of how to
  1530. 56:27build like my own version of it. I also
  1531. 56:29wanted it to inventory what is being
  1532. 56:31built out so I can review it. This is
  1533. 56:33another thing. If you're using a
  1534. 56:35long-running agent, it will it can go
  1535. 56:37and build like tons of components, and
  1536. 56:39then you will have to reason out when
  1537. 56:41you're judging it. So, it is always a
  1538. 56:43better idea to inventory it um so that
  1539. 56:46you can go and verify stuff. Because,
  1540. 56:48look, verification is something that
  1541. 56:50agent
  1542. 56:51cannot do. Like, it it can do in some
  1543. 56:53way, but it still needs like, you know,
  1544. 56:55your your judgment here. And that is
  1545. 56:57very critical now than ever.
  1546. 56:59>> uh I I'm curious, like, do you want to
  1547. 57:01take a minute to explain what is your
  1548. 57:04process to create the spec? And what's
  1549. 57:06your process to evaluate the what the
  1550. 57:08agent builds? Because, sometimes we do
  1551. 57:11feel when building is so easy, it can
  1552. 57:13feel slightly lazy to go and uh verify
  1553. 57:16the work. I do feel that sometimes.
  1554. 57:18Yeah. And manual verification does not
  1555. 57:20scale at the rate in which agents can
  1556. 57:23build things, right? So, what's your
  1557. 57:24process to evaluate uh whether the right
  1558. 57:26thing has been done or not?
  1559. 57:28>> Yeah, that's a good question. So, here's
  1560. 57:31the thing. I I I still think for
  1561. 57:33enterprise applications, right?
  1562. 57:36Verification is definitely
  1563. 57:39needs to happen at human level. So,
  1564. 57:43especially for enterprise because
  1565. 57:44there's a lot of things at stake. Now,
  1566. 57:46this is also a judgement to begin with
  1567. 57:48like if you know that this thing hasn't
  1568. 57:52that it will not have that many impact,
  1569. 57:54then you can move really faster, right?
  1570. 57:56Like sometimes I'm not even reviewing
  1571. 57:58the code. I'm like, "Yeah, okay. What
  1572. 57:59about the agent did? Let's just test it
  1573. 58:01out." Like smoke test, it's working
  1574. 58:03fine.
  1575. 58:04But sometimes
  1576. 58:06if it's very critical, right?
  1577. 58:08It depends on the degree of criticality
  1578. 58:10to answer in short, right? If it's very
  1579. 58:11critical, I I generally ask agent to
  1580. 58:15like do some queries and you know,
  1581. 58:17automate that way like you know, query
  1582. 58:19it through this data lake objects. And
  1583. 58:22if I have a smoke test data given, I
  1584. 58:25will sort of manually do or at least
  1585. 58:29manually keep an inventory that, "Okay,
  1586. 58:31if this thing happens, this is
  1587. 58:34you know, this is what I'm expecting,
  1588. 58:35right?" And then I ask agent to go and
  1589. 58:37run that loop for me. So, it will go and
  1590. 58:39run that loop, right?
  1591. 58:41So, I can trust what the agent did is
  1592. 58:44exactly correct or not.
  1593. 58:46>> Yeah.
  1594. 58:46>> Uh
  1595. 58:47but somebody like
  1596. 58:49it's still somebody has to keep the
  1597. 58:51system
  1598. 58:53in their head like entire head like
  1599. 58:55because see, it's like if you let agent
  1600. 58:57sort of take over that thing from you,
  1601. 59:00it will go and build a lot of things
  1602. 59:01that you have no idea and that's the
  1603. 59:03worst thing. Like whatever agent build
  1604. 59:06and if you don't have an idea about it,
  1605. 59:08that is the worst problem to have,
  1606. 59:10right? So, that's that's the shift that
  1607. 59:12is happening like because this
  1608. 59:14implementation is so these agents can
  1609. 59:16implement stuff very faster, right?
  1610. 59:18>> Yep. Yeah.
  1611. 59:19>> Um the the other part of the works
  1612. 59:21suddenly become so much more important,
  1613. 59:23right?
  1614. 59:24>> Yes.
  1615. 59:24>> Um yeah. So, so that's the thing
  1616. 59:27um that I would say. And uh yeah, I
  1617. 59:30think the agent is still kind of working
  1618. 59:32through it. So look what it did. So
  1619. 59:36here's the remaining new net mapping.
  1620. 59:38Let me get the the field it's trying to
  1621. 59:40call. But I don't know I expect a
  1622. 59:43possibly category I'll execute it.
  1623. 59:45Cannot map engagement. You see it tried
  1624. 59:48to map to the wrong thing and it
  1625. 59:51suddenly realized it is not the right
  1626. 59:53thing to do. Now remember like you know
  1627. 59:55at this point that is the human signal
  1628. 59:57is so much more important because for
  1629. 59:59the next iteration I can put this in the
  1630. 1:00:01skill. In the skill file. Like whenever
  1631. 1:00:04you are mapping these type of objects go
  1632. 1:00:07map to this, go map to that. So now you
  1633. 1:00:10know instead of like your point right?
  1634. 1:00:12It will go back and sort of
  1635. 1:00:15figure it out somehow.
  1636. 1:00:17These agents are goal driven. You have
  1637. 1:00:19given it a goal. They're going to figure
  1638. 1:00:21it out one way or the other. They're
  1639. 1:00:23going to reason keep reasoning. You know
  1640. 1:00:25like as you said you know it might have
  1641. 1:00:26wasted ton of tokens here doing some
  1642. 1:00:29unnecessary work but it is coming to the
  1643. 1:00:30same conclusion here. See it was like
  1644. 1:00:33contact point email mapping was created
  1645. 1:00:35it did create it did create some
  1646. 1:00:37mappings here. It's moving on.
  1647. 1:00:39But my point is same like see it created
  1648. 1:00:42the mappings now. You see
  1649. 1:00:44it went and you know
  1650. 1:00:46figured out that it needs to map to the
  1651. 1:00:48contact point email. So we have a common
  1652. 1:00:51email now, right? Now it's going to
  1653. 1:00:53create another object. Because it needs
  1654. 1:00:56that review object and it's going to map
  1655. 1:00:58the other fields to that. And then once
  1656. 1:01:00it's ready we'll go create the identity
  1657. 1:01:02resolution rule. So as you said like you
  1658. 1:01:04know sometimes the recipe is in your
  1659. 1:01:06head and you want to just communicate to
  1660. 1:01:08the agent. Sometimes you just let the
  1661. 1:01:10agent loose let it figure it out you
  1662. 1:01:11know and both of these worlds have their
  1663. 1:01:13own you know advantages
  1664. 1:01:16and their own drawbacks as you said like
  1665. 1:01:18drawbacks and pros and I think that is
  1666. 1:01:20where the human judgment suddenly
  1667. 1:01:22becomes like so important, right?
  1668. 1:01:25Uh yeah, so it's going to work out. I'm
  1669. 1:01:27going to leave my laptop, um you know,
  1670. 1:01:29and uh as you can see, like these things
  1671. 1:01:32can take for ever, like, you know, a a a
  1672. 1:01:35couple of hours. One of the settings I
  1673. 1:01:38did with my laptop is not to sleep, you
  1674. 1:01:40know. This is the first change that
  1675. 1:01:41happened after these agents and I saw
  1676. 1:01:44how they work, right?
  1677. 1:01:45>> Yep.
  1678. 1:01:45>> So, they can work and I think that is
  1679. 1:01:47where the cloud loads are also going to
  1680. 1:01:49become more critical, like, you can push
  1681. 1:01:51this workload to the cloud and let all
  1682. 1:01:53these things, you know, sort of happen
  1683. 1:01:55in the cloud. I can see that uh thing
  1684. 1:01:58happening as well. Uh in fact, not
  1685. 1:02:00happening. It's uh it's already
  1686. 1:02:01happening. Uh with that, actually, I
  1687. 1:02:03want to move back to the slides,
  1688. 1:02:04actually. I mean, this is going to
  1689. 1:02:06happen. Um I did run, as I said, with
  1690. 1:02:09that spec and it it built end-to-end
  1691. 1:02:11everything, which was surprising. Um
  1692. 1:02:14anyway, so with that, right? Uh I think
  1693. 1:02:16we are running short of the time. I
  1694. 1:02:18don't want to take much of your time as
  1695. 1:02:19well, Anush, but for all of you, right,
  1696. 1:02:21joining today, a big thank you, right?
  1697. 1:02:24Thank you for joining us and I hope you
  1698. 1:02:26learned and saw this MCP. Now, one of
  1699. 1:02:29the other things I want is feedback.
  1700. 1:02:31Definitely, that feedback is there. Um
  1701. 1:02:34take a screenshot of this one. I'm going
  1702. 1:02:35to keep it for a minute here. Um
  1703. 1:02:38this is all the resources that we have.
  1704. 1:02:40I'll also try to do a trailblazer
  1705. 1:02:42community post sharing all these things.
  1706. 1:02:45Um here we have blog post where we
  1707. 1:02:47introduce this DX360 MCP server. Um
  1708. 1:02:51there is another code live sessions and
  1709. 1:02:54um
  1710. 1:02:55uh that is not the code live session. I
  1711. 1:02:58think this session This is on-demand
  1712. 1:03:00video that we have to learn about code
  1713. 1:03:02extensions. I think they are also great.
  1714. 1:03:04You can ask agent to write code to do
  1715. 1:03:06certain aspects, like, rack pipelines or
  1716. 1:03:09if you're building um transformations.
  1717. 1:03:12Um and then, um you know, if you want to
  1718. 1:03:15learn about cloud code, how to build um
  1719. 1:03:18agent skills with Cloud Code, there is a
  1720. 1:03:20link to there where we talk about that
  1721. 1:03:23aspect of how to use Cloud Code to sort
  1722. 1:03:25of build your own own skills. So, all of
  1723. 1:03:28these resources are there. And then,
  1724. 1:03:30yeah, of course, um you know, the other
  1725. 1:03:32other aspect here I want you to remind
  1726. 1:03:34is um the Data Blazer community, right?
  1727. 1:03:38Um please post all your questions.
  1728. 1:03:40Anusha's there as well.
  1729. 1:03:42>> Yeah, I'll be there to answer if you got
  1730. 1:03:44If you guys have any questions from this
  1731. 1:03:46session, you want some follow-ups, you
  1732. 1:03:48want resources that you have not gotten
  1733. 1:03:50yet, me and Moit as well as our
  1734. 1:03:53engineering team will be in the Data
  1735. 1:03:55Blazer community, and we'll be able to
  1736. 1:03:57help you out. You can start a post about
  1737. 1:03:59today's session, ask any questions, and
  1738. 1:04:02we'll do our best to make you
  1739. 1:04:03successful.
  1740. 1:04:05>> Awesome. Awesome.
  1741. 1:04:06Yeah, with that, I think um you know, we
  1742. 1:04:09are end of this. And again, uh you know,
  1743. 1:04:11for all of you, um I I hope this was
  1744. 1:04:14educational. And uh I hope you're going
  1745. 1:04:17to play with the MCP. Please let us know
  1746. 1:04:20if there is anything else we can help.
  1747. 1:04:22Any parting thoughts here uh from you,
  1748. 1:04:23Anush?
  1749. 1:04:24>> Uh no, this has been great, Moit. I
  1750. 1:04:26think uh there's nothing better to use
  1751. 1:04:29or learn about the MCP server than a
  1752. 1:04:31hands-on session, right? This is not
  1753. 1:04:33something you want to hear people just
  1754. 1:04:34talking about it, but actually showing
  1755. 1:04:36how it works. And
  1756. 1:04:38I think you've done a great job in
  1757. 1:04:40educating all of us in what your
  1758. 1:04:42workflow was before, how it's changed
  1759. 1:04:44now, and the way things are moving. So,
  1760. 1:04:47uh I'm really glad. Thank you for
  1761. 1:04:48everyone joining you, and thank you for
  1762. 1:04:50Moit as well for conducting this session
  1763. 1:04:52for us.
  1764. 1:04:53>> Yeah, absolutely. Absolutely. And uh
  1765. 1:04:56yeah, um with with that, right? Um
  1766. 1:04:59again, a lot of you are asking in the
  1767. 1:05:00chat about how can we share some of
  1768. 1:05:03these um skills and open-source work
  1769. 1:05:06that as open-source work, um
  1770. 1:05:09we'll definitely make all the possible
  1771. 1:05:11attempts here to give you the
  1772. 1:05:12documentation and the skills. Um and
  1773. 1:05:16also the MCP server though is live, you
  1774. 1:05:18know, make use of it. Connect it to your
  1775. 1:05:20coding agent if you're working in data
  1776. 1:05:22360 space.
  1777. 1:05:24Uh your developer edition all or also
  1778. 1:05:26has all of these access. So, if you're
  1779. 1:05:28just learning, please make use of it. Um
  1780. 1:05:31and yeah, um you know, um
  1781. 1:05:33we will make sure that we will give you
  1782. 1:05:35all the all the resources
  1783. 1:05:37uh and always
  1784. 1:05:38please feel free to reach out to us.
  1785. 1:05:41And with that,
  1786. 1:05:42thank you. Thank you everyone.
  1787. 1:05:43>> All right.

About this transcript

This page contains the full transcript of Agentforce NOW codeLive: Build Agentic Data 360 Experiences Headlessly by Salesforce Developers, generated from the public captions YouTube serves with the video. The transcript has 11,682 words across 1,787 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.