Precisely, there are about 100 of these, and everyone makes a new one every week.

MCP is a dead end, just ignore it and it will go away.

> but we created MCP. Now we're wrapping MCP into CLIs...

Next we'll wrap the CLIs into MCPs.

MCP only exists because there's no easy way for AI to run commands on servers.

Oh wait there's ssh. I guess it's because there's no way to tell AI agents what the tool does, or when to invoke it... Except that AI pretty much knows the syntax of all of the standard tools, even sed, jq, etc...

Yeah, ssh should've been the norm, but someone is getting promoted for inventing MCP

Absoultely.

I'll add to your comment that it isn't a bug of MCP itself. MCP doesn't specify what the LLM sees. It's a bug of the MCP client.

In my toy chatbot, I implement MCP as pseudo-python for the LLM, dropping typing info, and giving the tool infos as abruptly as possible, just a line - function_name(mandatory arg1 name, mandatory arg2 name): Description

(I don't recommend doing that, it's largely obsolete, my point is simply that you feed the LLM whatever you want, MCP doesn't mandate anything. tbh it doesn't even mandate that it feeds into a LLM, hence the MCP CLIs)

Yup, routing is key. Just like how we've had RAG so we don't have to add every biz doc to the context.

I agree with the general idea that models are better trained to use popular cli tools like directory navigation etc, but outside of ls and ps etc the difference isn't really there, new clis are just as confusing to the model as new mcps.

Yes we just RAG to be applied on tools. Very simple to implement.

Token counts alone tell you nothing about correctness, latency, or developer ergonomics. Run a deterministic test suite that exercises representative MCP calls against both native MCP and mcp2cli while recording token usage, wall time, error rate, and output fidelity.

Measure fidelity with exact diffs and embedding similarity, and include streaming behavior, schema-change resilience, and rate-limit fallbacks in the cases you care about. Check the repo for a runnable benchmark, archived fixtures captured with vcrpy or WireMock, and a clear test harness that reproduces the claimed 96 to 99 percent savings.

The AI prose is getting so tiring to read

"We measured this. Not estimates — actual token counts using the cl100k_base tokenizer against real schemas, verified by an automated test suite."

For me (actually trying to get shit done using this stuff) it's validation.

Being able to have a verifiable input/output structure is key. I suppose you can do that with a regular http api call (json) but where do you document the openapi/schema stuff? Oh yeah...something like mcp.

I agree that mcp isn't as refined as it should be, but when used properly it's better than having it burn thru tokens by scraping around web content.

The point is authorization. With full web access, your agent can reach anything and leak anything.

You could restrict where it can go with domain allowlists but that has insufficient granularity. The same URL can serve a legitimate request or exfiltrate data depending on what's in the headers or payload: see https://embracethered.com/blog/posts/2025/claude-abusing-net...

So you need to restrict not only where the agent can reach, but what operations it can perform, with the host controlling credentials and parameters. That brings us to an MCP-like solution.

A few things: in this case, you have to provide the tool list in your prompt for the AI to know it exists. But you probably want the AI agent to be able to act and choose tools without you micromanaging and reminding it in every prompt, so then you'd need a tool list... and then you're back to providing the tool list automatically ala MCP again.

MCP can provide validation & verification of the request before making the API call. Giving the model a /tool/forecast URL doesn't prevent the model from deciding to instead explore what other tools might be available on the remote server instead, like deciding to try running /tool/imagegenerator or /tool/globalthermonuclearwar. MCP can gatekeep what the AI does, check that parameters are valid, etc.

Also, MCP can be used to do local computation, work with local files etc, things that web access wouldn't give you. CLI will work for some of those use cases too, but there is a maximum command line length limit, so you might struggle to write more than 8kB to a file when using the command line, for example. It can be easier to get MCP to work with binary files as well.

I tend to think of local MCP servers like DLLs, except the function calls are over stdio and use tons of wasteful JSON instead of being a direct C-function call. But thinking of where you might use a DLL and where you might call out to a CLI can be a useful way of thinking about the difference.

One thing that I currently find useful on MCPs is granular access control.

Not all services provide good token definition or access control, and often have API Key + CLI combo which can be quite dangerous in some cases.

With an MCP even these bad interfaces can be fixed up on my side.

The prophecy of the hypermedia web

Proxying / gatekeeping

When you're using an agent, the "query" isn't just each bit of text you enter into the agent prompt. It's the whole conversation.

But I do wonder about these tools whether they have tested that the quality of subsequent responses is the same.

Basically yes.

Skills or tell it the --list command would be my guess.