Ask HN: Why do we need MCP?
10 points - last Saturday at 10:26 AM
Why not just publish your API docs and let the client's bot chain together what calls need to be made.
You could even repurpose the MCP so you only ever produce the API documentation as EVERY response. No LLM necessary.
poxlr
last Saturday at 9:47 PM
MCP and APIs are extremely similar, but MCP provides a common protocol that everyone can follow, so every agent can use the same protocol regardless of whether the backend uses REST, OpenAPI, or GraphQL. It will just work, and agents know how to crawl other MCPs; therefore, providing an MCP tool makes it much easier for an agent to develop a usage pattern.
hackeryogi
yesterday at 4:34 PM
To add to this, MCPs usually focus on capabilities. This is especially useful in multi-hop situations.
As an illustration, lets say I ask the agent to “Read the issue Bob posted today from Linear (issue tracker), pull relevant logs from DataDog (log tool) and suggest a fix”
In the MCP world, my agent doesn’t have to spend tokens / time understanding APIs or the protocol (as the parent comment suggests). Instead, it relies on the tool provider for the best way to “read the issue” or “pull logs”.
Same thing, in the API world, would usually involve multiple API hits (and possibly using the wrong APIs or params for them), interpreting the results and hitting few more APIs again.
alex7o
yesterday at 5:19 PM
My biggest problem is that msps can't be chained together and that the response schema is not know models are very good at filtering with jq to get only the relevant fields
vpk11
last Saturday at 10:53 AM
I use CLI tools for Git, Atlassian (ACLI), and Metabase, plus custom SigNoz shell scripts to drive API calls instead of an MCP server. This setup works very well at the moment for me.
20Cprogrammer
yesterday at 1:31 AM
I run a few agents side by side and I've been going back and forth on whether to bother with MCP servers at all. I'm not clear on what each of those is doing in your loop though. Could you go into a bit more detail?
VladTkDev
last Saturday at 5:18 PM
[flagged]
jwally
last Saturday at 10:28 AM
Probably covers most use cases and is a lot cheaper and safer.
Antolius
last Saturday at 10:53 AM
Having a dedicated spec makes it easier for MCP client apps to implement some nice features out of the LLM loop (e.g. OAuth flow or elicitation ), some handy user facing features (e.g. prompts and resource), and some more advanced use-cases (like server sent notifications).
But yeah, a lot of it (including tools) maps well to HTTP semantics. In fact, at work we implement our features as HTTP APIs, and we’ve built a framework that exposes them as MCP servers.
legalwarrior
last Saturday at 10:50 AM
Whats an mcp ? Mcp connectors Iv heard of
al_borland
last Saturday at 11:23 AM
From my limited understanding, it’s basically an API, rewritten for us by AI agents. So the humans can consume your REST API, while the agents will use the MCP to effectively do the same thing.
owebmaster
last Saturday at 10:56 AM
Is anyone forcing you to use MCP? You don't need it. If you can't see the benefit of it, don't use it.
jwally
last Saturday at 11:02 AM
It's why I asked the question. Maybe I missed something.
owebmaster
last Saturday at 11:18 AM
MCP is an API. The biggest difference is that it defines some primitives that AI clients know how to implement beforehand so every MCP server works similarly, otherwise each API implements its own way to list the tools available, change/react to changes, authorization, session management, etc. MCP *is* an API.
jwally
last Saturday at 11:29 AM
I'm thinking more as a provider. Why put a model on my side to handle permissions/scope/auth/actions when thats already handled (or easy enough to spin up)by a rest-inspired http api?
If you're needing to give the model direct and broad access to your SqlServer instance...I don't get it.
Dedicated customer facing API is like whitelisting approved actions; MCP feels like getting an LLM to enforce a complex blacklist.
owebmaster
last Saturday at 1:44 PM
Check WebMCP that ChatGPT Desktop just implemented for a good example.
Now every website can create tools that gpt can use. So it's possible to have your users use AI without the need to pay for tokens.
AznHisoka
last Saturday at 11:09 AM
What’s the harm in being curious and seeing what you might be missing?
owebmaster
last Saturday at 11:14 AM
There's no harm in being curious but why frame it like the thing you don't know how to use is not useful? That's harmful.
vasyl_kyryliuk
yesterday at 8:18 PM
[flagged]
cjzhao
yesterday at 2:57 AM
[flagged]
dsk-dev-ai
yesterday at 9:32 AM
[flagged]
gytlr01
last Saturday at 3:16 PM
[flagged]
dgotlieb
last Saturday at 12:33 PM
[flagged]
laruss5
last Saturday at 10:40 AM
[flagged]
MITC
last Saturday at 7:16 PM
[dead]
bsramin
last Saturday at 1:22 PM
[flagged]
htmxsucks3
last Saturday at 2:47 PM
[dead]
speedgoose
last Saturday at 1:37 PM
MCP is more or less API and discovery/documentation. It’s very far from being the first technology to do that.
What it does well though is to keep things relatively simple. It is also good for things that takes time to return.
I find it competitive also for non LLM usages. I have done some PoCs with MCP instead of REST/OpenAPI, or GraphQL, or Protobuf and I liked MCP.