I Built GTM Grid: Run GTM Workflows From Claude Code

0:00This is one of the first multiplayer

0:03GTM style tools that I've come across.

0:05And yesterday I posted a video around

0:09the new Buzz Slack feature, which is a

0:11multiplayer agent kind of workspace that

0:13is an open-source version of Slack. And

0:16this effectively is an open-sourced GTM

0:19tool that allows you to do so many

0:22different things. Think about it like

0:24NA10 combined with Clay, combined with

0:27you know, large data sync tools. It's

0:30really, really cool. And it's totally

0:31hooked up to your Claude code, your

0:34Codex, your Cursor, or whatever harness

0:37that you want to to use. And so it's

0:39completely controlled. Everything inside

0:42of this is totally controlled by the CLI

0:45and the MCP that it creates. So, I

0:47wanted to go over a use case that I'm

0:49just kind of fleshing out right now.

0:52So, the first thing to mention, and

0:53again, this is where it's again, totally

0:55different. There is like a cloud

0:56version. There's also just an

0:57open-source version where you could

0:58self-host it. But it has this sync

1:01ability, so you can effectively sync, as

1:03you can see here, I'm syncing my Atio

1:05into the grid here, and we have all of

1:08the people coming into into this. And

1:11what we can start to do is we can start

1:12to build a lot of cool things off the

1:14back of it. First of all, we could do

1:16some various elements of automations. Of

1:18course, it runs kind of like a a kind of

1:21like table-style grid anyway. So, it

1:24does kind of more of a rudimentary

1:26workflow, which is sequential like

1:28column by column. But what you can also

1:31do is you can create this thing in here

1:32called pipelines. Pipelines is

1:34effectively where you then can start to

1:36run various if nodes, loop nodes to run

1:39more complex automations inside of a

1:42column. So, you could have different

1:44columns, then inside of one column it

1:46could be this particular pipeline that

1:47you create, which then spits out another

1:49result as well, which is

1:51uh really cool.

1:53So, the other thing that I really like

1:54about it is like one of the things that

1:56I want to do now is update or make sure

1:58that my CRM stays totally up-to-date.

2:00So, I'm going to go into plan mode. We

2:02can use different models here. I'm just

2:03going to use 4.8. It's fine, but you

2:05could go on to Fable or or Opus 5, I

2:07guess.

2:08And we're going to say, I want to ensure

2:10that my CRM stays up-to-date.

2:14Um how would we go about checking the

2:18current person still works at that

2:22company? If they don't, I want to then

2:26go ahead and update the person record.

2:30So, what it's going to do now, and

2:31again, this is So, when you first sign

2:33up, it [clears throat] just connects to

2:34your local Claude code that's been

2:35running in the terminal. It's really,

2:37really simple. If you If you're only

2:38using the app, you just open up a

2:40terminal and you do Claude login, login,

2:42and then it basically finds you've

2:44logged in and hooks up to that.

2:46Um same thing for Codex and and

2:48effectively the same thing for Cursor.

2:50Pretty cool as well, because you can go

2:52ahead and you can obviously use tons of

2:54different models that they have within

2:56there, as well.

2:57But, I I do tend to just use Claude

2:59inside of here. Now, at the moment I'm

3:00using plan mode. I really, really like

3:02this, because before I go and like make

3:04massive edits to a table, I want to go

3:08ahead and work on the entirety of the

3:12plan for the table. So, really work with

3:13the AI to flesh all of this out. And you

3:16can see here, it's basically going and

3:18querying the table. It knows which table

3:20you're on, because it can see it

3:21basically here. So, depending on which

3:23table you go to or pipeline it will

3:26basically have access and can view that.

3:28You could also tell it to go to another

3:29table and it'll be able to go do that,

3:31which is really cool. It's a really cool

3:32kind of co-pilot um experience here. And

3:35so, it's created the plan. So, let's

3:36just view the plan before we approve it.

3:39So, here it's saying, okay, uh

3:42relevant columns name, email address,

3:44company, job title, LinkedIn.

3:45Fine, fine, fine, fine. So, the first

3:47thing is to check uh change check add a

3:50job change check function. So, here it's

3:52going to use lead magic. So, obviously

3:54it has access to tons of different tools

3:57and obviously can use custom HTTPs and

3:59stuff like that. It's then going to

4:01return if they've changed. Then it's

4:04going to do a kind of code-based node,

4:07which is effectively saying that they

4:08have moved or not. It's going to verify

4:11before it does a bulk run and then it's

4:12going to go ahead and use the Atio asset

4:15record up upsert person. So, basically

4:18what it's then going to do is it's then

4:19going to go ahead and update the

4:21individual information inside of the

4:24Atio through an Atio node as well. And

4:26we could actually set this to run every

4:28single day to check as well as we can

4:30set like a daily cron, which is really

4:32cool. And then I'm going to say I want

4:34to actually add to this plan. I also

4:39want to check every single I'll check

4:43let's say once a week on the latest

4:46posts within these people. Check that

4:50they have posted about intent data or

4:54signals.

4:56If they have, I want that to be added

4:59into Atio as a note plus a Slack

5:03notification to me. Okay. So, now it's

5:07going to then basically go ahead and

5:08create the plan associated with that.

5:11And then it's as simple as just

5:12accepting the plan if you're happy with

5:14it. It's going to tell you which tools

5:16it's using. It obviously has These are

5:18the AI providers that you can use as

5:19columns inside of here, so you can use

5:21open router anthropic API. Uh your local

5:24Hermes agent, which I found really cool

5:26for using like copywriting because

5:28obviously my Hermes agent has so much

5:30information logged in there, so using it

5:32inside of here has been massive. And

5:35then obviously all your tools. The

5:36really cool thing is you can see who's

5:38on it. When your agent starts working, a

5:39little agent kind of bubble comes up

5:41here and you can watch what it's doing

5:43as well. So, when my team has been using

5:45this, I can see everything that they're

5:47doing as well. And there's this really

5:48cool kind of like multiplayer experience

5:51where they can tap into everything that

5:52everyone's doing as well. So, it's been

5:54really fun to build out all of our

5:56campaigns through this tool, and it

5:58makes everything so seamless, quick, and

6:00easy to do. Now, you don't even need to

6:02do it through this app. You can

6:04literally just sit on your Claw Code app

6:05or your terminal, and it has access

6:07through to the MCP, so it can do it

6:08through there as well. But, to be

6:10honest, like, you know, the reason that

6:11I really like this tool is I really like

6:13to have some level of visualization when

6:15I'm doing you know, some copy-based

6:17writing. I really like to look at, you

6:19know,

6:19what the code is doing from a

6:22go-to-market's perspective. So, they

6:23kind of add this kind of UI interface.

6:25So, let's go have a look at the plan in

6:27a little bit more detail. So, we've got

6:28the part one, which is the check.

6:32Then part B, weekly post signal. So,

6:34here it's going to use Triggerfy, and

6:36it's going to use Triggerfy to fetch the

6:38social media posts. Then it's going to

6:40use AI to basically check if they

6:42mention intent data or signals, which is

6:44cool. Then it's going to branch if it's

6:47true. And then it's going to use the

6:49Atium node and then a Slack node to send

6:51an email. And then we can literally just

6:53say,

6:54approve and run. And this thing is now

6:56going to go ahead and start to build out

6:58everything with that we're going to do.

7:00And you just now got to let that cook

7:01for, you know, whether it takes 1

7:03minute, sometimes it takes longer if

7:04it's a big thing, up to 10 minutes or

7:06so. But, either way, it's just going to

7:07go ahead and cook for you as well. So,

7:11really cool tool. Go check it out. It's

7:12totally open sourced with some kind of

7:14paid features stacked into there. I've

7:16started to contribute tons to this as

7:18well, and building on top of it. And I

7:20just think it's a really cool product.