Author: Brian T. Jackett

MCP, Microsoft Graph, and Copilot connectors in GitHub Copilot agent mode

Brian T. Jackett MCP, Microsoft Graph, VS Code , , , , , , , , ,

[Update 2025-07-09] After internal discussion with Merill Fernando and others, it does appear that you can use delegated authenticated with interactive auth prompt. I will be testing this out soon and share more once testing is complete. [/Update]

Recently I had the chance to test out a number of new technologies as part of my org’s Fix Hack Learn (FHL) week. This includes an MCP (model context protocol) server for Microsoft Graph to ingest public data into a Copilot connector using GitHub Copilot agent mode. While I’ve had lots of experience with Microsoft Graph and Copilot connectors (formerly Graph connectors), the other areas were new to me. In this post I’ll walk through my experience and observations based on my time spent so far.

In general, I was pleasantly pleased with how easy it was to get started. It literally took 5-10 mins starting from nothing to having my first successful prompt successfully responding from the MCP server. The following days I dug in really deep and got a bit lost trying out different models, adding instruction files, and testing out various prompts. By the end of the week, I had a working process that you can see in this shortened video.

Use case

To frame the scenario, my goal was to identify a set of publicly available data (ex. data.gov) and ingest that into M365 Copilot via a Copilot connector without writing any code. Once the data is ingested I could then use M365 Copilot to reason over the data through user prompts.

In my reading thus far, MCP is generally used as a way to expose “tools” to an LLM so that the LLM can identify APIs or other resources to use when processing a user prompt. In my case though I was essentially treating MCP as a means to prototype a development artifact (i.e. Microsoft Graph requests that worked with users, Copilot connectors, SharePoint sites, etc.) Think of this like a developer REPL akin to Microsoft Graph Explorer but instead of calling APIs directly I used natural language prompts as the input.

Getting started

This post is not intended to be an introduction to MCP, but for those interested I found the following resources useful.

https://modelcontextprotocol.io/introduction
https://www.anthropic.com/news/model-context-protocol

From the Microsoft Graph perspective, there is not an official Microsoft Graph MCP server, but I did find a community-based MCP server called https://lokka.dev from my peer Merill Fernando. You can run Lokka locally which I did as an NPM package in VS Code (now natively supported) as well as other options.

Merill provided a quick and easy guide for installation and running the Lokka MCP server. A few of the set up instructions didn’t match the UI screens as this is recently GA’d and subject to change, but I shared with Merill and he was able to update as of the time of writing.

https://lokka.dev/docs/install

Ensure that you have the pre-requisites installed. In my environment that includes (may be higher than stated pre-reqs):

  • Node v22.16.0
  • VSCode v1.101.2 (user setup)

Lastly, ensure that you have created and consented permissions on a Microsoft Entra application registration. Note that at this time, as far as I’m aware, this currently only supports application (i.e. app-only) permissions. There are some workarounds I’ve read about to implement delegated permissions, but it doesn’t appear to be formally supported or part of the MCP spec as of yet. I was originally led to believe that only application permissions are supported (which is what my sample used), but after internal discussion I’m told that delegated auth is now available although not widely adopted by many MCP servers yet. I will be testing delegated auth over the coming week or two and share a separate update.

I purposely only consented the following permissions so that I could test out scenarios that weren’t consented:

  • User.ReadWrite.All
  • ExternalConnection.ReadWrite.OwnedBy
  • ExternalItem.ReadWrite.OwnedBy

Comparing LLM Models

One thing that was not clear to me at first was that GitHub Copilot agent mode supports different LLMs, but more importantly that some of those models are considered “premium” models and have different consumption / usage limits. Additionally, some models (ex. Claude Sonnet 3.7) are only accessible with a paid plan such as Copilot Pro or similar.

https://docs.github.com/copilot/get-started/plans-for-github-copilot

I tried out three different models and landed on GPT-4.1 as my preferred, but recommend that you compare models to best suit your solution needs.

https://docs.github.com/copilot/reference/ai-models/choosing-the-right-ai-model-for-your-task

https://docs.github.com/copilot/concepts/copilot-billing/understanding-and-managing-requests-in-copilot

Ex. screenshot of differing consumption rates depending on model used:

Testing out commands / models

I attempted a number of simple scenarios such as getting users from the tenant, filtering on user properties, and updating metadata on a user. All of these were successful with varying formats of output depending on the model used.

  • GPT-4.1 was my preferred model, primarily because it struck a good balance between figuring things out on its own, good formatting, and prompting me as the developer before proceeding with various creative / destructive actions.
  • Claude Sonnet 3.7 attempted a number of actions on its own without asking me for input and in many cases those has undesired effects. This was the “magic” that I referred to in a previous LinkedIn post. In just 2 testing prompts, I had created a Copilot connector, registered a schema, and ingested sample items. Sadly the schema and sample items were very far off from what was intended and thus this soured my experience.
  • o4-mini on the other hand was very sparse on information that it returned and I perceived it to be much slower (i.e. time to first token in response was at least 50-100% slower than the others).

GPT o4-mini (preview)

Claude Sonnet 3.7

GPT-4.1 [my preferred so far]

I ran a number of other scenarios as well that are too lengthy to share here, but these include:

  • Creating a Copilot connector (see video near top for overview)
    • Scraping information from website pages
    • Using instruction files (for GitHub Copilot agent mode) to encourage / discourage prompting developer for confirmation of actions
  • Attempting to access resources not granted permissions to
    • ex. read SharePoint files, access Exchange mailboxes
  • Modifying user data in Microsoft Entra
    • ex. assign job title, department, email address, etc.

Conclusion

I am very happy with my time and outcomes testing out MCP + Graph + VS Code, but realize that this is not a production-worthy solution yet. First among them is only supporting application permissions, many of which require global admin consent. There are still big gaps when it comes to fully supporting delegated OAuth, fine-tuning the right amount of control vs. creativeness of the models, and rough edges getting a scenario implemented. I see lots of potential and will continue to keep an eye on this space.

If you have any questions or suggestions for additional use cases, or are already using MCP in your solutions, please share in the comments below.

-Frog Out

Retrospective for 2024

Brian T. Jackett Personal

Continuing my tradition of writing a retrospective (2023, 2022, 2021, 2020, 2016, 2015, 2013, 2012, 2011, 2010) or look ahead (2019, 2014, 2013, 2012, 2011, 2010), in this post I’ll recap 2024 and share goals for what is ahead in 2025.

Disclaimer: Yes, I know it is already 3+ months into 2025 as I write this. I’ve put off writing in general for too long, but it’s never too late to start back up again.

Milestone

Last year I turned 40. It feels odd to say that in general, but looking back over the last ~2 years it has been a time of big change personally, professionally, and with my family. I recently commented on a Bluesky post from a random user who mentioned they were also turning 40.

Going from 39 to 40 was a much bigger "reflection -> take action" point in my life than I would have imagined. I had a list of really big "someday I'd like to…" items that I started consciously working on and now I'm actively pushing forward with in my life.Make time for yourself.

— Brian T. Jackett (@briantjackett.com) 2025-04-10T17:12:04.437Z

When I was 39 and looking ahead to 40, there were a number of big things that I wanted to do in the coming 5+ years. I think that arbitrary age of “turning 40” was a good motivational kick to get started on those goals. Already I’ve started:

  • Learning Spanish (again) – 755 day streak on Duolingo
  • Training for half marathon race
  • Started Diaconate pre-req classes and formation discussions
  • Applied and accepted into management position for my team at work

If you are looking at big, audacious goals in your life, what concrete steps are you taking to achieve them and on what timeline?

Exercise

As mentioned above, one of my goals is to complete a half marathon race. In 2023-2024 I increased my walking routine on my home office treadmill, getting up to 12 miles per week (4 days a week of ~3 miles). Now that we’re into 2025 I’ve also registered for a 5k (completed in March), 10k (2 months away in June), and half marathon (Oct 2025).

https://www.linkedin.com/embed/feed/update/urn:li:share:7309914046776913920?collapsed=1

Recapping 2024 here were my stats for the year.

https://twitter.com/BrianTJackett/status/1742557691359203655

  • 194 days with exercise
  • 371 miles ran (400 total with cycling)
  • 160.3 hours ran

A big thank you to a number of friends who have shared training plans, nutrition advice, and encouragement for my half marathon training. I’m also incorporating other strength and core workouts to help prepare. Looking forward to crossing that finish line in 6 months!

Reading

Since 2021 I have been reading books more regularly. You can view my profile on GoodReads along with “read” and “want to read” lists.  I set my 2024 GoodReads challenge to read 8 books and ended at reading 9. A few were last minute short reads at the end of the year, but still made it. One thing I did notice is that I had gotten very heavy into religious books (for classes and related) but found that I enjoyed mixing up into other topics. Going forward I’m trying spread out topics so that I keep up interest in reading more regularly.

A few noteworthy books from 2024:

  • The Power of Showing Up
    • My 3 children are starting to grow into adolescence / pre-teen. My wife and I have had conversations about making the most of our time with our children while they are still young and making memories. I appreciate how this book gives simple advice for “being present” with family members. It is not about what you buy them, lavish trips / vacations, or even the words that you say to them. It is about simply showing up and creating an environment where children feel safe and encouraged. I would recommend this to parents / caregivers, but primarily with younger children if possible.
  • The Return of the Prodigal Son: A Story of Homecoming
    • As a Catholic / Christian, I’ve heard the story of the Prodigal Son dozens of times for decades. I read this book as part of a men’s group at my church. I was fascinated with how the author dissected various aspects of this painting, applied them to different perspectives of the characters in the story, and related it to his own personal life and journey. Even if you are not practicing a faith, highly recommend reading for reflection.
  • Rest: Why You Get More Done When You Work Less
    • Many of us can likely relate to feeling overwhelmed, not getting enough rest, and constantly playing catch up after you do “rest”. I enjoyed the ways this author outlined building rest into your daily / weekly / annual life, how to use rest time more effectively, and setting expectations around self-care. You can’t perform at your best if you are constantly fatigued. Personally I’ve struggled with this (borderline burnout) in the past year+, so this is definitely an area I need to work more on.

For my 2025 GoodReads challenge I’ve set a goal to read 8 books again and already making progress.

Diaconate formation

As mentioned in my 2023 retrospective, I’ve begun pre-requisite classes to become a deacon within the Catholic Church. I have completed 5 requirements and only have 1 remaining. I’m already starting to read the book for the next class so that I don’t get overwhelmed with class work in addition to work / family life. I’ve also been helped by meeting every other month with a spiritual director (priest from my hometown) to discuss life, faith, and my future endeavors.

Work & management role

In early 2024, I applied for and was accepted to take a management position for my current team (M365 Copilot Extensibility CPx). I had been discussing with my then manager about moving into a management role. I had been interested to take a small step back from the technical focus of my role and instead grow in guiding the careers of my teammates. It has been just over a year since I took on this role and it has been very rewarding so far. I can’t say if this will be my continued path forever, but I look to continue pushing myself and others to be the best that we can.

One aspect that has been a challenge in the last year is the pace of work life combined with my children growing older and being active in more activities. As I tell some of my peers, the AI boom has been “gas pedal to the floor for 12-18 months” and not let up much. In the past we used to have cycles of intense pushes and then lulls to build up for the next push. With our AI advancements, most of those former lulls are now additional mini-marketing moments, pivots to more promising focus areas, or similar. I truly treasure the time off that I take for family vacations or other needs. It is important to “touch ground” so that I can recharge my own personal batteries.

Learning Spanish

One interesting side item is that I started to (re-)learn Spanish. In grade school and high school when I took 5 years of Spanish classes. I signed our family up for Duolingo (everyone now learning at least 1 language / music / mathematics) as well as try to find time to practice with others from my church (large Hispanic population) or peers at work. I’m on a 760-day streak so far and enjoying this very much. If you’d like to practice together sometimes reach out and let me know.

Conclusion

2024 (and now part of 2025) feels like it has flown by. Reflecting on what I wanted to accomplish has truly been helpful in making intentional choices on how I spend my time, deciding what I say no to (to make time for the things I say yes to), and setting up longer term plans. I hope 2024 treated you well and that you are making progress on your own goals. Feel free to share any of your experiences in the comments or link to your own reflection from 2024.

-Frog Out

Credits
Coin operated telescope – Matt Howard
Pair of blue and white adidas running shoes – Sporlab
HOLA LED signage – Jon Tyson

Presentation on Copilot Extensibility to Copilot Learning Hub

Brian T. Jackett Copilot for Microsoft 365

In September 2024 I joined Dona Sarkar on the Copilot Learning Hub video series to discuss M365 Copilot extensibility via declarative agents, Graph connectors, and more. Great opportunity to share out more about how customers and partners can extend Microsoft 365 Copilot with additional knowledge, custom actions, and more!

https://learn.microsoft.com/shows/copilot-learning-hub/unlocking-microsoft-365-copilot-extending-functionality-with-external-data-and-integrations

-Frog Out

Search and Retrieve Content for Microsoft Graph Connector Items via Graph APIs

Brian T. Jackett Office 365, Search

In this post, we’ll walk through a scenario for surfacing Microsoft Graph connector items via the Search APIs on Microsoft Graph and then using values in the response to retrieve the full content of item using the Graph connector APIs on Microsoft Graph.

Scenario

This week, a customer asked the following question:

“Once we index files using Graph connectors, how can we retrieve the full content of those items to be used in additional experiences like a custom search solution for our employees? The Search APIs (on Microsoft Graph) appear to only retrieve a snippet of the matching results.”

As the customer noticed, when you use the Microsoft Search API to search Microsoft Graph connectors, you can list which fields (properties from Graph connector defined schema) to return, but you can’t return the full content that was ingested. This is by design as search is intended to retrieve snippets of results for displaying in a custom search experience, for example.

Solution

Let’s look at querying for Graph connector items using the /search/query endpoint (documentation):

Note: replace “ServiceNowKB1” with the id of your connection.

POST: https://graph.microsoft.com/v1.0/search/query
REQUEST BODY:
{
    "requests": [
        {
            "entityTypes": [
                "externalItem"
            ],
            "contentSources": [
                "/external/connections/ServiceNowKB1"
            ],
            "query": {
                "queryString": "*"
            },
            "from": 0,
            "size": 3
        }
    ]
}

Sample response from Search API:

Screenshot of sample response of Graph connector items using the Search API through Graph Explorer.

Notice the “substrateContentDomainId” value in response. You can parse the value and take the substring after the comma (,) such as highlighted above. This is the itemId of the Graph connector item.

Next, query the Graph connector ingestion API (documentation) on Microsoft Graph for the items in the Graph connector (externalConnection). Again, replace “ServiceNowKB1” with the connection id as well as replacing the itemId.

GET: https://graph.microsoft.com/v1.0/external/connections/ServiceNowKB1/items/c487857187032100deddb882a2e3ec4f

Sample response from Graph connector API:

With the ingestion API we are now able to retrieve the full content of the ingested item. Note that this is for a single item at a time. If you have multiple items, you will need to iterate through them.

Conclusion

Graph connectors power discovery of non-Microsoft 365 data in multiple native Microsoft 365 experiences. If you have a need to display Graph connector items in an alternate location (line of business application, custom search portal, etc.) hopefully you found this post helpful. Feel free to leave comments or questions if you have additional needs.

-Frog Out

Resources

Header photo

Presenting at Microsoft 365 Community Conference 2024 in Orlando, FL

Brian T. Jackett Conference/Tech Event, Copilot for Microsoft 365, Presentation

Join me on Apr 30-May 2, 2024 as I’m presenting “Understanding Microsoft 365 Copilot Extensibility Via Microsoft Graph Connectors” at the Microsoft 365 Community Conference. Registration is still open and there are pre/post workshops also available. Great to see so many subject matter experts sharing on topics related to Microsoft 365, Copilot for Microsoft 365, Azure, Dynamics, and more. My specific session will be on Wednesday May 1st, 2024 at 10-11am ET. Look forward to seeing you there!

Use this attendee discount code to save $100 USD off registration: MSCMTY

Register today!
https://aka.ms/M365Conf24

-Frog Out

Presenting at Copilot for Microsoft 365 Tech Accelerator

Brian T. Jackett Copilot for Microsoft 365, Presentation

Join me on Feb 28-29, 2024 as I’m presenting “Expanding knowledge with Graph connectors” at the Copilot for Microsoft 365 Tech Accelerator event. This free, online event will have numerous subject matter experts sharing on topics related to Copilot for Microsoft 365. My specific session will be on Feb 29, 9-9:30am PT. Look forward to seeing you there!

Register today!
https://aka.ms/AccelerateCopilot

-Frog Out