MCP Hello Page(hybridlogic.co.uk) |
MCP Hello Page(hybridlogic.co.uk) |
This feels like less of a hack and more of discovering what some of the HTTP headers are for. You’re choosing rather reasonably how to present the resource found at /mcp when a client is asking for the resource to be presented in HTML format. It’s perfectly fine to offer an HTML response that says “hey this is not really presentable in HTML. Do this instead.”
The specification still leaves a lot to be desired, especially as it relates to auth. There are lots of bad ways to do auth with MCP and only a couple of good ways. It also puts a lot of pressure on the various IdP vendors and relies on lesser used areas of OAuth 2.0/2.1 (like DCR, token exchange, etc.). It started out in a place where the assumption was you were running an MCP Server on a laptop or you were a SaaS provider serving lots of individual users -- somehow DCR in the initial spec iteration seemed like a good idea (spoiler: it wasn't) and fortunately, the latest revision has somewhat addressed that. XAA/ID-JAG & CIMD should continue to round-out some client management and auth solutions for the enterprise.
Gateways are another area that needs to be addressed in the spec. There isn't a formal definition of one in the spec, and yet, there are lots of "gateways" out there. What a gateway is and what it should do is an open question and it means different things to different people depending on who you ask. For example, who does token exchange: the MCP server or the MCP gateway? Both are valid answers right now depending on the implementation or your opinion of which is best.
More spec iterations should be coming this year. I'm still pretty optimistic about MCP as a whole, as it remains a good way to standardize tool calls across agents and some of the other entities that it provides like resources and prompts are genuinely useful to add more determinism to an agent. Interceptors and skills will be good, too.
If you're interested in helping to evolve the spec further, the MCP Contributors Discord is active. There are lots of IGs/WGs that solicit feedback and you can participate in meetings with your feedback.
I have also been finding the MCP auth story to be really lacking was excited to see OAuth 2 support until I tried to get it to work at work and realized our idp implementation didn't support 2.1, and went into the spec and started wondering if anyone had a good experience yet. Luckily most of our environment can settle on a OAuth token env var standard until that's all in order.
We’re excited about XAA, it will simplify many flows
"Sorry, you have been blocked
You are unable to access hybridlogic.co.uk
Why have I been blocked?
This website is using a security service to protect itself from online attacks. The action you just performed triggered the security solution. There are several actions that could trigger this block including submitting a certain word or phrase, a SQL command or malformed data."
I think this probably also helps when truly clueless users drop the link into an agent directly, because then the agent will relay the message to the user.
i have learned that if the request accepts html, just give them that, however, some agents will ask for text/markdown, so just give them that.
you can even take OPs ideas further: text/html? serve html help page; text/markdown? serve the same but pure text, maybe nudge the agent on how to install the mcp; application/json? pass through to the mcp app.
the whole llms.txt thing could have been that
Is this not the intended use of request headers?
Spec-based implementations of OAuth 2.0/2.1, including dynamic client registration and token exchanges, are absolutely necessary on a large scale but are an enormous barrier for adoption during early deployments. Cookie tricks ignore the spec and produce special cases when client credentials need to be rotated.
Solution in the real world: configure the server to provide the WWW-Authenticate header with value "Bearer" on unauthorized requests, issue limited-scoped tokens, leave the rest up to mcp-remote. The client authenticates once, uses bearer token in its requests, server verifies against scopes. No dynamic client registration required.
The authentication solution becomes clearer if you distinguish between "what the spec allows" and "what you need in your deployment." Not many servers really require OAuth on day 1.
I once worked for $COMPANY and we had a network scanning application. Always generated a lot of tickets from angry people wanting to scream about bots.
So we put a web page page on each worker that would inherit some details from whatever job was running, and say “I’m a $TYPE_OF_SCANNER FROM $COMPANY doing $THING_THAT_BENEFITS_YOU.
This behavior is covered by our terms of service page at $LINK.
If you believe that we should not be doing this, please contact $SUPPORT and provide this code:
$SCAN_JOB_IDENTIFIER”
Call volume and unhappy customers went way down.
ex: Googlebot/2.1 (+http://www.google.com/bot.html)
You can still keep using Rest API with swagger docs and tell the AI to read the swagger docs. It's the same thing. The entire Rest API specification is a lot more flexible than the JSON RPC format /mcp uses.
Personally, for just tools, I wire up tool calls with context captured via system prompt.
ideally what you would have is skills + scripts that the agent can load progressively to interact with the api
or even better: a single cli tool the agent is aware of, and which combines `--help` output with `tldr` tips, when the command is incomplete, and prints out helpful error messages when the incantation is wrong
In fact, this is how Kubernetes APIs work. Even the CLI asks YAML or JSON content type when `-o yaml` or `-o json` is specified.
Also, some other sites (albeit sometimes looking at the User-Agent header, hence more hacky) also doing the same. For example `ipinfo.io` returns HTML page when opened in a browser but a JSON (prettified) when requested via cURL.
There are indeed some limitations to this implementation. The primary one being IPv6 support. The implementation prioritizes convenience over internet limitations, requiring us to roll out IPv6 dual stacking on the web request level as opposed to the DNS level. On an API level, this results in users with IPv6 addresses making API requests to v6.ipinfo.io.
So, last year we rolled out dedicated API infrastructure for api.ipinfo.io
GET / accept: text/html -> passthrough to existing content
GET / accept: text/markdown -> respond with llms.txt or a stripped down version without html tags or other noisy bits
Can I just say that anybody involved with MCP's launch should be ashamed of what they put out there. I understand tool calling. I understand specs. I read MCP's "spec" and I used useless word salad that alternates between baby's first wire format and pie-in-the-sky marketing speak. Several of the navigation links I encountered were broken.
Poorly thought out, poorly communicated, but it's the only thing out there that 1) meets the need, and 2) published by people with a huge amount of reach. Right place, right time, shit effort. So it gets adoption. Like the history of PCs, of the internet, or everything, I guess. Worse truly is better.
https://github.com/modelcontextprotocol/modelcontextprotocol...
For example, is a typical Google Drive user expected to install an official google drive CLI and a skill md file just to interact with Google Drive in the chatgpt app? That seems absurd, IMO. Clicking “install” via an MCP / Plugin marketplace and then logging in via Oauth is less confusing.
MCP is also way more convenient (as a user) if i’m interacting with an API that supports oauth anyway. I don’t need to generate an api key, install a binary, create an md file, etc. CLIs and skills are great too! But not always the best choice.
note: cli tools do auth too — gh and claude code itself have some of the nicest auth flows i've experienced
CLIs are working better with current harnesses in my experience, but no idea how I would ship them.
some folks go to the trouble of publishing with `brew`
then there's the old `curl -LsSf https://example.com/install.sh | sh`
mcp is basically `claude mcp add <service name> <service url>`
skills require multi step marketplace registration, refresh, then picking the right skill out of *checks notes* 116 skills on the official marketplace + however many skill come with the new marketplace
cli tools may be a bit better, maybe they're there already because you need them to interact with deployment, or they might be a simple `brew add <service cli>`
I like their solution. It feels like because remote MCP servers are built on HTTP that they are actually using the spec as intended. Serving html when asked for it.
Isn't this literally against the entire point of MCP
you only deploy + maintain one mcp server, all mcp capable harnesses will know how to use it.
you need to publish to each marketplace for discovery, but you can also just put the the incantation on the product page and be done with it.
claude mcp add --transport http <name> <url>
codex mcp add <name> --url=<url>