Hi all,

I have some questions about working with Claris MCP and Claude Desktop. 

Q1 : How do I add additional tables to an existing connection in Claris MCP? The documentation says I can add connections in Editing state, but my interface only shows 'generate configuration' with no option to add or edit connections.

Q2 : How do I prevent Claris MCP from returning large container/media fields in update tool responses? The full record data exceeds 1MB and causes Claude errors, but tool descriptions and schema parameters don't control the response size.

Q3 : When using OData vs Data API with Claris MCP, OData returns all fields including large container fields that exceed 1MB, whereas Data API allows managing this through special layouts. What's the recommended approach to handle large container fields in Claris MCP - should I use Data API instead, or is there a way to configure OData to exclude container fields?

Q4 : When setting up Claris MCP connections, how should I choose which tables/relationships to include from a large FileMaker relationship graph if editing the connection later isn't available? What's the best strategy for selecting items initially without excluding tables I might need later?

Q5 : Claris MCP Developer Tools Issue - Simple Summary

Problem: Documented Developer Tools for connection management are not being exposed by the Claris MCP server.

What We Need: Access to connection management tools (specifically update_filemaker_connection) to add tables to existing connections without deleting them.

Current Status:

Context is in Editing state ✅

Claude Desktop configured with "Always available" tool access ✅

Only 7 tools appear (6 connection-based + 1 developer tool)

Connection-management tools completely absent ❌

Missing Tools:

get_connections

add_filemaker_connection

update_filemaker_connection

remove_connection

Other connection management tools

Impact: Cannot add tables to existing connections or modify connection configuration as documented.

Question: Is there any way to access the connection management Developer Tools - through Claude Desktop, terminal, API, or any other method?

I am looking thru this now...

https://help.claris.com/en/claris-mcp-help/content/index.html

Hey man, interesting questions - and quite a few of them. So, my thoughts....
A lot of what you're running into isn't necessarily you...its the current state of the controlled release.

As far as adding tables to an existing MCP connection. I believe the documentation is a little bit ahead of the product, at this point. the UI only exposes generate configuration and then the wizard - right? The editing state existsa, but the UI does not surface the connection-management tools that the docs reference - YET. I THINK.

There is still more to come with FM26 and agentic coding.

As for preventing the large container fields from being returned - try using the Data API instead of oData. I can help you connect to that if needed. Create a special layout that omits the container fields, only the stuff you want.
Point the MCP's data api connection at that layout. I do a lot of work with the data api, and it's both nice and annoying to have to populate a layout with the specific fields you want. Hope that helps.

OData returns all fields, including containers
Data API respects the layout-based visibility. Hopefully, this gives you some ideas. I am writing a connection tool right now that i'm thinking of releasing to the community (if people want it) that makes switching between odata/data api pretty easy. And exposes the connection to web. Whether or not you created your app agentically.

As for the missing developer tools. I am fairly certain this is another example of the docs being ahead of production. It's simply functionality that Claris has not enabled yet - but should be coming!

You aren't doing anything wrong. MCP is just a limited early release. I do not think it is a finished product, and at [Full Access] that was made pretty clear. Big things are coming. Claris is really pouring their heart into moving the platform forward. Hope that helps a bit. Feel free to message me if I can help in any way. I've posted on Claris's forums, but never made an account here.

Hi Alex - thanks for your quick response

When I created a new context and then created a new connection, the app came up with these various defaults...

Your FileMaker account must have both the OData (fmodata) and Data API (fmrest) extended privilege sets enabled. The Data API is used internally for table and script discovery during setup.

Yes, I have both the both the OData (fmodata) and Data API (fmrest) set up for this file

---

Search Contacts records - description box

And as an example, it added this by itself...

Searches Contacts records using FileMaker OData syntax (100 records default).

Q: - are you telling me that somehow I could switch this over to use data API instead?

And if so, I imagine I would have to include a layout in this description so how would I update this?

Thanks in advance for your help

You seem like you're right on track brother. MCP sort of auto generates those default descriptions when you spin up a new connection, I believe.
It'll pull whatever it can on setup. That's likely why you are seeing the 'Searhc contact records using FM Odata syntax" thing.

Yes you can absolutely switch that over to the data API. MCP doesn't expose a beautiful UI for it, yet, buit the underlining connection is still just pointing wherever you want to point it at. The trick is making a utility type layout in FM that has only the fields you want the agent (or yourself, or whomever) to see. Then you point the Data API connection at that layouty

$layout = "Contacts::Barebones";
Something like that in php. Whereas barebones has only what you need.

What I do, when trying to get enough data on those layouts, is lean on PORTALS! THey're a godsend. You can get all kinds of info on your utility layouts without having to manually do it all and portals come through the data api as portalData, I believe. versus data/fmResponse[0]['data'].

Id have to pull up a sample response to give you the correct order of the object/array.
Right now, the part that’s missing is the ability to edit the connection after the fact. That’s why you’re stuck with whatever MCP generated on day one. The docs talk about update_filemaker_connection, but those tools aren’t surfaced yet. So you aren’t doing anything wrong. It’s just the current state of the controlled release.

If you want to switch that “Search Contacts” tool over to Data API, you basically recreate the connection using Data API instead of OData, and make sure your layout is set up cleanly. Once Claris exposes the editing tools, you’ll be able to tweak these without rebuilding anything.

If you want, I can try to walk you through setting up the Data API version when I have some time, or else you could always just hire us as consultants. I don't mind helping if i can though. I’ve been doing a lot of this lately and it’s pretty straightforward once the layout is dialed in.

when you get use to working with both of them - it becomes second nature to design with these situations in mind. I've learned so much from 2 years of working primarily with data API and integrations. Little tricks and tips you'd never know. Hope that helps.

Alex Seidler
[email protected]

Hi Alex,

Thanks So much for your excellent help!

In my case, these are the default settings when you create a new connection. And as you can see everything says odata...
So my question is if I was going to try to convert this over to data api - how could I do that?

The description box says this

Searches for records in the Contacts table using FileMaker's OData syntax. This tool will return 100 records by default.

Could I switch it over somehow and say this?

Searches for records in the Contacts table using FileMaker's Data API syntax. This tool will return 100 records by default.

And then how would I tell it to use a specific layout?

Search for records using layout "MySimpleLayout"

And then, at the same time we have the input schema

And as you can see, it is currently set to odata...
The OData API $filter operators. The expression specified for each record in the collection. Items evaluated as true are included.

Anyway, this is all quite interesting.

And if I had to use odata for searching, and I wanted to eliminate container fields - would I just tell it the Fields I wanted it to return?

Returned only these fields - Firstname, Lastname, Company, Phone .... something like that?

Thanks so much for thinking about this