All articles
Solution guideNov 3, 2026

How to connect Claude, ChatGPT or Copilot to your field operations with MCP

If your team already works in Claude, ChatGPT or Microsoft Copilot, it can ask Sytex about tasks, sites, forms and projects from the same chat. You type “Which overdue tasks does my team have?” and the assistant looks it up in your Sytex organization and replies with the list and a link to each task.

The connection uses the official Sytex MCP server. It runs on Sytex’s own servers, so there is nothing to install or build. You click the button for your assistant, sign in with your Sytex user and pick the organization.

What is an MCP server, and what does it change in field operations?

MCP (Model Context Protocol) is the open standard AI assistants use to connect to other systems. When an assistant asks for an “MCP server” or a “custom connector”, it is asking for the address of the system it should read from and act on.

In field operations, that system is where the work is tracked: sites, each crew’s tasks, the forms technicians fill in on site, materials and milestones. Without a ready server, connecting an assistant to that data means building and maintaining your own integration. Sytex already runs one, and it answers with what your team has logged up to the moment you ask.

What can you ask Sytex from Claude, ChatGPT or Copilot?

The assistant can look up tasks, projects, workflows, forms, sites, materials and dashboards, summarize how a project or a workflow is going, and search your organization’s Knowledge Base. It can also create or update records, such as changing a task’s status or assigning it to someone else. You write the way you would write to a colleague:

  • “Which of my tasks are overdue, and what is blocking them?”
  • “Summarize the progress of the Project North project by workflow.”
  • “Which forms were submitted this week and are still waiting for approval?”
  • “Create a task called Site survey in the Project North project for next Friday and assign it to me.”

The answers are as current as what the field team logs in Sytex. If that work still lives in spreadsheets and chat groups, this guide on moving field work out of Excel and WhatsApp covers the step before.

Is it safe to connect an AI assistant to your field data?

The assistant works as your Sytex user. It sees and does only what you can see and do in Sytex, with your same permissions.

  • You sign in on a Sytex page, with your password or your company’s single sign-on (SSO), and authorize the assistant by name. The session renews on its own.
  • Each connection works in one organization, the one you picked when you authorized it. To work in another, add a second connection with another name, for example Sytex Operations.
  • Claude and ChatGPT ask before changing anything. They show you the record they are about to create, edit or delete and wait for your approval.
  • A Copilot Studio agent does not ask for confirmation before it creates, changes or deletes data in Sytex. Ask it for specific changes and check the result in Sytex.
  • You can remove access at any time. In Sytex, click your profile picture, choose View profile and click the plug button, Connected assistants. The access is removed right away.

How do I connect Sytex to Claude?

It works on the Claude Free, Pro, Max, Team and Enterprise plans.

  1. Click Add Sytex to Claude. Claude opens the Sytex connector.
  2. Click Connect, sign in to Sytex, pick the organization and click Authorize.
  3. In a conversation, click + in the message box, go to Connectors and turn on Sytex.

On Team and Enterprise, a workspace owner enables the connector for the organization first. Then each person clicks Connect with their own Sytex user.

How do I connect Sytex to ChatGPT?

You need a ChatGPT Plus, Pro, Business, Enterprise or Edu plan. On Business, Enterprise and Edu, the workspace admin has to allow the app first.

  1. Click Add Sytex to ChatGPT. ChatGPT opens the Sytex app.
  2. Click Connect, sign in to Sytex, pick the organization and click Authorize.
  3. In a new conversation, click + next to the message box and select Sytex.

How do I connect Sytex to Microsoft Copilot Studio?

You need permission to create or edit agents in Copilot Studio. You add Sytex to the agent as an MCP server, and the agent can then answer in Teams or Microsoft 365 Copilot. If your company restricts connectors with data policies, IT has to allow the Sytex one.

  1. Open your agent, go to Tools, click Add a tool › New tool and choose Model Context Protocol.
  2. Fill in Server name with Sytex, a description such as “Looks up and updates Sytex tasks, projects, forms and sites”, and https://mcp.sytex.io/mcp as Server URL (https://mcp.eu.sytex.io/mcp if you open Sytex at an address ending in .eu.sytex.io).
  3. Under Authentication, choose OAuth 2.0 with type Dynamic discovery, then click Create and Next.
  4. Choose Create a new connection, sign in to Sytex, pick the organization the agent will work with and click Authorize. Back in Copilot Studio, click Add to agent.
  5. In the Sytex tool’s authentication settings, set Credentials to use to End user credentials. Each person then signs in with their own Sytex user and the agent works with their permissions.

Frequently asked questions

Does Sytex have an official MCP server?

Yes. Sytex hosts it, and it works with Claude, ChatGPT, Microsoft Copilot Studio, Claude Code and Codex.

Do I need to build my own MCP server to ask ChatGPT or Claude about my field operations?

Not if the work is tracked in Sytex. The server is already running, and each person connects with their own Sytex user and permissions.

Which Claude or ChatGPT plan do I need?

Claude works on every plan. ChatGPT needs a Plus, Pro, Business, Enterprise or Edu plan.

Can the assistant change data in Sytex without asking?

Claude, ChatGPT, Claude Code and Codex show you the change and wait for your approval. A Copilot Studio agent does not ask, so request specific changes and check them in Sytex.

See how it works in Sytex: Rollout management for network operators · Field app for contractors.

Back to blog
Webinar transcriptFull text of the video (21 min)

0:00Let's go. Hello, everyone. Good afternoon. Good morning, wherever you're joining us from. I'm Mariano Lisiotti. I'm the CEO and co-founder of Sytex. And today, we're starting a series of webinars where we'll show you how to integrate your favorite AI agents into Sytex. Sytex is a project management tool that helps teams deliver their projects faster. And now, with the help of Claude and also OpenAI's models, we're showing how you can really automate everything that's part of a project, and hopefully help you deliver projects faster.

0:53So today is going to be very quick. We're going to present a couple of slides just to give you a brief idea of what it's all about, and then go into how to do it: connecting Claude and OpenAI's agents to Sytex. Great, let me see how I can share my screen. There we go. Also, I'm not sure if I can see comments or questions. I think you can leave them in the comments section of the event.

1:46So this is the agenda for today. What's an agent, and how is it different from a chat? What are MCP and skills, and then a demo. I think it's four or five slides. What's the difference between ChatGPT, let's say, and an agent? With a chat, you just type a question and it goes away and brings back your answers. It can search the web, or it can bring information from what the model already knows, and answer questions with that. An agent, and in Claude that would be Claude Code or Cowork, you have Codex on OpenAI's side and other tools from other models.

2:38An agent is a model where you give it a goal and it goes ahead and uses the different tools that are available to it to get that work done. Ideally, if you have an agent installed on your computer, like Claude Code, it will also use the files on your computer, different tools you have on your computer, and also tools that are available to that model and that you gave it access to on the web. So now let's talk about how agents connect to Sytex. My take is that the AI models or companies are putting those two concepts together.

3:29I think OpenAI made a move this week where chats and agents are going to start looking the same, and you probably won't notice a difference. But for now, if you're using Codex, Claude Code or any of those agents, this is the way to configure and connect Sytex to them. Now let's move into how to connect Sytex to these agents. And that's through a connector called an MCP. There you go. Thanks for the insight. So the MCP is the way an agent talks with another platform, in this case with Sytex, to understand what it does and hopefully do some activities together.

4:25So we just launched an MCP connector that's available in Claude and also in OpenAI, and it's very easy to configure, and that's part of the demo we're doing today, where you can do pretty much anything you do in Sytex through the MCP, and it will use your user, your permissions, all the security measures you have in your account to use Sytex, just as if you were using it yourself in the interface. So another concept... so it's clear now, right? We have Sytex on one end, the agent on the other, and the MCP connects both worlds.

5:12But there's a missing piece, because if we want to do something in Sytex, we don't want to be explaining all the time to the agent what we need it to do. Imagine we want to review the quality of a document, or launch a new process, or get a summary of what's going on in our projects. We don't want to explain the whole process over and over again to an agent. That's where skills come into play. A skill is a way to help the agent understand a procedure or how we want things done in general, right? But what's the problem?

5:57What's the catch with skills today with these large language models? That our skills are kind of a local thing, stored on your computer. And Sytex is a project tool focused on teams collaborating on information. So we created a tool in Sytex called the Knowledge Base, which, if you're a user, you'll see in your main menu, where teams can put their own skills. And through that, share common knowledge on how to do things. You can create any type of skill and use it as a repository of all the information you have and that you want your team to use, and also iterate on it across all the actions and all the agents running in your organization, right?

6:52So hopefully that makes sense. Now let's go to the demo. Again, feel free to ask any questions in the comments, and we'll try to tackle them at the end. Great. We have Claude here, which has two use cases for this. This is the regular chat, and here we're using Claude Code. Make sure you stay on Claude Code, or if you use Cowork, go through that. You can see it's running on this computer, and you can connect your file system to it.

7:40Or a GitHub repository, that's more for the coders out there. But what we're going to do here is connect the Sytex MCP to Claude. So we'll click here and go to Settings. And down here, we'll go to Customize. And we'll add a new... we'll search for Sytex, sorry. We'll search for Sytex. The new Sytex connector shows up, right? I could explain a bit of what it does, but you just click Connect to link Sytex and Claude. And what it does... it came up on another screen. This screen shows up, and we click Connect. Sorry, this is in Spanish for some reason.

8:27But now it goes into Sytex, and it tells me: an app is asking for access to your Sytex account, and it says which app it is, in this case Claude Code, my user, and now which instance, which organization I'm going to give access to. I'll just click a demo organization I have. Hopefully, on your side, you'll have one or two organizations you belong to. Now I'll click OK, and that's done. That's how the whole thing works. Now Claude Code is connected to the Sytex MCP, and we can start playing around with it.

9:13So I can come here and type: what are my pending tasks in the Claude Webinar project in Sytex? It'll go find the tools. You can see what it's doing there. It's using Sytex, finding out how to do things. The connector explains to the agent how Sytex is used and what it does, and all the concepts it needs to go and find that data on the platform. Great, let's give it a moment... yes, there you go. So here we have... I should be able to hide this, yes, there you go.

9:59So here we have the list of activities we have for this project, some permits, some network design, some tasks due today. I can click on them and it opens Sytex and shows me the task and everything I could do to complete my activities and keep working. Let's do something else. Let's go to OpenAI. Here we are.

10:40I'm accessing it from the web. If you don't use Claude or Anthropic and you're using OpenAI, ChatGPT, the process is very similar. Try clicking Work here on the main screen. And then when you go to the settings, you have a menu here called Plugins. And in Plugins, just type Sytex and hit Enter. Oh, sorry, Browse directory. There you go. Sytex, and the connector shows up. Again, the same process asks me to validate access.

11:32I'll give it access to the demo instance again. And we're done. So now ChatGPT is connected to my Sytex account, and I can go ahead and ask questions. So I'll type the same question I just typed in Claude, and it'll go ahead and do the work. Actually, I had it on high, I should probably lower that to make the demo faster. Great, so that's step number one: connecting the MCP to your agent, and through that, your Sytex account. The second thing we wanted to do here is work on skills, right?

12:18So I could give Claude a command here. I have all these texts typed here on the side. Give me a status report for the Claude Webinar project in Sytex. It'll create some kind of report, thinking about the structure, with the structure the agent finds useful for me. But let's say we want a specific report that I want to request every week or every month to be sent or shared across different teams. To do that, I'll create a new skill for Claude to store in Sytex, and then we'll use it for reporting, right?

13:15So again, I've typed here a very short definition of a new skill. While it's thinking, I'll show you how it's done. Again, this is very simple. You can just tell it: create a skill in the Sytex Knowledge Base called Weekly project status report, with these rules. So it's like a little traffic-light status, which tasks are overdue, and rules for how to show different scenarios depending on the status of the task. See, the report came back from Claude, and it gave us a structure that's okay, but it's probably not what we want or what we're used to.

14:01So with this, I'll ask it to create the skill in Sytex. And skills, again, are a great tool to teach the model how you want things done. And the cool thing is that all the skills, through the MCP connection with Sytex, are available to the agent all the time. And the agent gets to pick which skill it wants to use, based on a very short summary that skills have at the top of the document. So the more you use skills and update them and keep them shared with your team in Sytex, the better the agent will work, on your computer and across your different teams.

14:58Let me open Sytex here to show you how that looks. There you go. Let me check. I created a weekly project status report. Nice. Let's find it. I think I can click here. Yes, of course. There we go. So we have this weekly project status. What we see here is just the summary, and that's the part the agent sees to decide whether this skill is useful. You can also see the actual instructions it created based on my input, and you can adjust them if you want.

15:48This is all live in Sytex, and available through permissions and different access levels if you want to keep that level of security. And with this skill, I can now go back and ask it for a new report, right? Can I just repeat it from here? No, that's not what I wanted. Let's delete this. And let's go back and ask again. There you go. We'll type that, and hopefully it'll now use the skill we shared to present the report again.

16:48Let's give it a few more seconds to finish. We could also use... where is it? There we go. In the meantime, we're maxing out our tokens on both fronts here. I can also ask ChatGPT something. Let's see what Claude is doing. There you go. All right, see, now it's using the traffic-light scenario, and also identifying the blockers, the quality, and so on.

17:32And now I have my report the way I want it. Here it's suggesting adding submitted forms to the skill to improve it, right? If I click Yes, update the skill, it will automatically update the skill in Sytex to keep it up to date and improving with every change. The last thing I wanted to cover... oh, we're 20 minutes in. So I just wanted to give you a quick look at the last thing I wanted to show, which is using the Knowledge Base we have in Sytex to present reports or information. So I can say: go ahead and create an HTML report, place it in the Sytex Knowledge Base, about this project summary you just made.

18:38So what it'll do... it'll probably take some time... but what it'll do is go to Sytex and create a nice artifact with the report's information. And you can see it in Sytex, and also share it with your different teams or different people, even outside Sytex. It's a great way to keep the repository of that information in one place. So actually, where is it? There we go. An example of that is the presentation we just did: it's an artifact, exactly that. We made it with AI, it's a presentation stored in Sytex, and I can share it with people on my team or publish it to the web, and it stays secure, and everyone will know where to find it and it's available to your team.

19:39Great, let's give it a couple more minutes to finish the task we assigned, but that was basically the presentation I wanted to share on how to connect Sytex to your favorite AI agents. We're going to keep doing this. There's a new webinar coming up on dynamic interfaces you can build with Sytex. And there you go, the report preview. Now let's see if I can open it.

20:25Well, it's still working. Yes, there's another webinar coming up. And of course, we're doing these webinars in three languages: English, Portuguese and Spanish, just kicking things off in English. But great. Hopefully it was useful. And we'll see everyone next time. You can reach us through LinkedIn. You can go to our website or to our social media accounts to reach out with comments, questions, or for help with any of this. Great, there you go.

21:12So that's it. For now, we'll see everyone next time. Have a good afternoon, good night, good morning, depending on where you are in the world. Thank you very much.