Many come to HN also for the comments
I think this biases towards sufficiently large engineering organizations where OAuth 2.0 was identified as necessary for some part of their requirements. In most organizations, they're still using `x-<orgname>-token` headers and the like to do auth.
I'm not sure that there's a better / easier way to do Auth with this use case, but it does present a signficant hurdle to adoption for those who have an API (even one ready for JSON-RPC!) that is practically ready to be exposed via MCP.
I don't think that's it. Auth is a critical system in any organization, and larger organizations actually present more resistance to change, particularly in business critical areas. If anything, smaller orgs gave an easier time migrating critical systems such as authentication.
https://vulnerablemcp.info/ is a good collection of the immediately obvious issues with the MCP protocol that need to be addressed. A couple low blows in there, that feel a bit motivated to make MCP look worse, but generally a good starting point overall.
If doing an extended, service-level session (like a GPT interaction) with a server known beforehand, would it make sense to set up a keypair and manage the interaction over SSH?
Restated: are we throwing away a lot of bandwidth establishing TLS trust for the more general HTTP?
nip•10mo ago
Excellent write-up and understanding of the current state of MCP
I’ve been waiting for someone to point it out. This is in my opinion the biggest limitation of the current spec.
What is needed is a tool invocation context that is provided at tool invocation time.
Such tool invocation context allows passing information that would allow authorizing, authentication but also tracing the original “requester”: think of it as “tool invoked on behalf of user identity”
This of course implies an upstream authnz that feeds these details and more.
If you’re interested in this topic, my email is in my bio: I’m of the architect of our multi-tenant tool calling implementation that we’ve been running in production for the past year with enterprise customers where authnz and auditability are key requirements.
jensneuse•10mo ago
Yoric•10mo ago
I mean, converting a tool-less LLM into a tool-using LLM is a few hundred lines of code, and then you can plug all your tools, with whichever context you want.
nip•10mo ago
My point is about the need for a spec of this mechanism: without a spec, every company / org will roll out their own and result in 500 flavors of the same concept.
That’s where MCP shines: tool calling and tool discovery is already 1.5 years old (an eternity in ai land).
The MCP spec ensures that we can all focus on solving problems with tool calling rather than wasting time in cobbling together services that re not interoperable (because developed without a common spec / standard)