mooreds

Enjoying this so far.

> The MCP authorization spec is written entirely around user-delegated flows and does not define a machine-to-machine grant at all.

is slightly wrong. There's a blessed extension: https://modelcontextprotocol.io/extensions/auth/oauth-client...

Yeah, the small bits of sand in the gears is something that is hard to sell when you allocating engineering effort. But it adds up over time.

It makes the difference between a tool that is a pleasure to use and one that causes dread.

This is my favorite article on the different ways to use AI.

https://danielmiessler.com/blog/keep-the-robots-out-of-the-g...

FusionAuth | Principal Software Engineer, Senior Java Engineer - Cloud, Account Executive | Varies between REMOTE (in USA, also in Europe but only for the account exec position) and ONSITE in Denver, CO, USA, details in each job desc | Salary ranges for the Principal Software Engineer it is 225k-270k, but the Euro positions don't have them :(

At FusionAuth, our mission is to make authentication and authorization simple and secure for every developer building web and mobile applications. We want devs to stop worrying about auth and focus on building something awesome.

There are a lot of companies in the auth space, but we feel like we have something special:

* a relatively unique deployment model (self-host on-prem, run in your cloud or let us operate it for you in ours)

* A well designed API first approach; one customer compared our APIs to petrichor

* a mature product (the code base is nine+ years old and we've found and fixed a lot of the sharp edges around core login use cases; but don't worry, there are plenty more features to add)

* a full featured free-as-in-beer version which makes the sales cycle easier; prospects often come in having prototyped an integration

Our core software is commercial. We open source much of our supporting infrastructure. Technologies and standards that you will work with: modern Java, PostgreSQL, Docker, Kubernetes, MySQL, OAuth, SAML, OIDC.

Learn more, including benefits and salaries, and apply here: https://fusionauth.io/careers/ ( Click/tap the 'View open positions' orange button. )

> it just makes the handoff between myself and the agent feel more seamless.

This terrifies me because there's no way for the services on the other side of the browser to know who is doing what.

How do you constrain the agent?

How do you keep track of what you have done vs the agent?

What am I missing? Am I just behind the times?

Edit, added reference to what terrifies me.

Thanks, I changed it from

Native American Tribes Came Together to Secure Their Rights to Colorado River Water.

to

Native American Tribes Secured Rights to Colorado River; States Are Stalling

That's a good point. The behavior varies wildly based on the domain provider and the behavior when you let a domain expire is similar to what happens when a phone number is deactivated, but with a possibly bigger blast radius.

You might find this podcast episode, where the topic is the economics of kidney transplants in Iran, interesting: https://www.econtalk.org/tina-rosenberg-on-the-kidney-market...

It's from 2015, so no idea if it reflects the current situation. But it digs into some of the same concerns (money, ethics, etc) you raise.

> As someone running a company, I get 2-5 unsolicited "vulnerability reports" per week. Half of them are an LLM finding some bad CSS on our framer splash page. The other half I assume are an extortion attempt so we just mark as spam.

I don't think that is unique to the LLM era. The company I work for has been getting some form of spam vulnerability reports years before LLMs were a thing. Often similar to what you mention about 'bad CSS'.

Maybe the volume has increased a bit, but we've added in a filtering solution and I'm more distant from the reports now, so hard to be sure.

Ooops, you are 100% correct. I will update the essay to use the correct term.

> It has to come with appropriate information, not just a date.

That's a great point. You need to give the manager enough info they can determine if it makes sense to intervene. Sometimes they have enough context that a date for a decision is all they need, other times you need to provide more details.

Haha, so much to learn so I wouldn't say I'm well-versed.

But let me know when you launch. My email is in my profile.

No worries, always appreciate the discussion. Good luck on the launch!

Have you seen WebMCP[0]? Had a customer ask about this recently.

It seems like it might be something of what you are looking for, since it leverages HTML to tell agents about website functionality.

One issue I see with WebMCP is that agents basically free-ride on user identity and authentication, which is problematic in some scenarios.

0: https://developer.chrome.com/docs/ai/webmcp

> you can technically do with tools too

By tools, do you mean skills? Or do I misunderstand you?

Thanks for sharing your solution. I clicked around and the information on your site is sparse. The postgresql page[0] doesn't really illustrate how your system works; just says that it does.

Regardless, there's a big difference between a proprietary way to inject skills and MCP, a standardized way to control access and deploy AI compatible logic.

0: https://adaptive.live/integrations/postgresql

> The problem with MCP is the initial setup friction is far greater than users login in themselves and grabbing the information they need.

Can you tell me more about this? With just-in-time client registration (DCR or CIMD) it seems like the MCP registration would be pretty simple.

Is it the configuration of the MCP client to know about the MCP server that is the issue?

Does the website need to be able to advertise "here's the corresponding MCP server" so that the "claude hits website" step becomes "claude hits website, discovers MCP server"?

Hiya, congrats on shipping!

Seems like the main use case is employees of companies. Is there an analogous use case/value for non-centralized users like customers or freemium users?

I'm struggling to think of one, but wonder what I'm missing.

Edit: I see you addressed this here: https://news.ycombinator.com/item?id=48594381

I think that this extension points out other benefits to MCP over skills:

   * centralized control
   * ease of use for employees
   * auditing/compliance
   * deployment model

It seems the state of the art for deploying skills is "copy this file and put it in this place" or "check out this repo and add a symlink" or "run this slash command to install the skill". (I'm not aware of any solution that pushes skills out.)

These options are simple, but not as easy as this extension makes rollout of a new MCP server to an employee.

Yeah, I read this and thought "makes total sense that Okta would be one of the prime movers behind this". The more I think about it, the more this is just like an app dashboard that an employee can use to access many different apps and that the enterprise can use to control and monitor access. So it makes total sense that they solved this (very real) problem.

I like one of the author's characterizations[0].

> Now employees no longer have to connect MCP servers manually and wait for a series of OAuth and login prompts. Once you log in to Claude from Okta, all the preconfigured MCP servers are already connected! It's not every day you get to improve both usability and security!

0: https://aaronparecki.com/2026/06/18/12/claude

Yeah, I recently documented how to use pre-registered clients with FusionAuth[0] (my employer). DCR's newer, better sibling CIMD is on our radar and under active discussion[1], but not currently available.

An alternative to the proxy you suggest is to generate a new Entra client id (with PKCE enabled) for every MCP client in a developer portal or similar, then have the user configure their client with that client id. Here's the CLI command I found to do this[2], but I bet there's an API too. Here are config instructions for Claude Code[3] and ChatGPT[4].

Client pre-registration is acceptable, but not optimal, for developers, and is a first class citizen in the spec[5]. If your main audience is internal and you can expect them to follow configuration instructions to get access to the MCP server, this approach can work.

But it's definitely not acceptable for widespread, public integrations if your audience is not developers. That is where a lot of the power and opportunity for MCP lies.

0: https://fusionauth.io/docs/extend/examples/controlling-acces...

1: https://github.com/FusionAuth/fusionauth-issues/issues/3230

2: https://learn.microsoft.com/en-us/cli/azure/ad/app?view=azur...

3: https://code.claude.com/docs/en/mcp

4: https://developers.openai.com/api/docs/guides/developer-mode

5: https://modelcontextprotocol.io/specification/2025-11-25/bas...

We actually state in our job listings that we will have an onsite interview as part of the process (we pay for travel/hotel, of course!).

Seems to weed these folks out.

Are there any products out there that are capturing the prompts/sessions? I imagine you could do it in an adhoc way, asking Claude to write up a summary of the session as part of the commit message. But is there anything else that's more structured/higher level?

Very similar, both let you decode and verify signatures of JWTs, with educational resources.

Basically, two sites that help devs understand JWTs, from two different companies.

> Is it some type of candy that looks like the fruit?

No, they are a variety of blueberries.

> Why is it on-topic?

It would have been better to post something about how they were bred, but I just ran across them on that site and wanted to post. They look, as you said, like candy, and I thought that was weird and interesting.

Edit, just found this from 2023: https://fruitgrowersnews.com/article/pink-cosmo-berries-a-hi...

How are agents authenticated?

I searched the docs for authentication and mcp (one of the protocols which, among other things, handles some pieces of authentication/authorization) but didn't see any results.

What did I miss?