Skip to content

AI tools · agent integration protocol

Model Context Protocol (MCP)

A widely supported standard for plugging tools and services into AI agents with built-in auth and scoped access, though critics note heavy token use and a rough, unevenly supported spec.

Summarized from 995 comments.

A widely supported standard for plugging tools and services into AI agents with built-in auth and scoped access, though critics note heavy token use and a rough, unevenly supported spec.

recommend
409
criticize
317
months recommended
12 of 13
median owned, 11 owners said
1 year

Why people love it

  • Broad client adoption, plug-and-play · 7

    MCP in its current direction and form is really an OAuth Protocol over http. And it has something other that other agent identity protocols don't: client adoption

    May 2026 · comment

    The whole purpose behind MCP was to provide a low-impedance standard where some set of tools could be plugged into an existing agent with no pre-knowledge

    Nov 2025 · comment

    There is huge value in having vendors standardize and simplifying their APIs instead of having agent users fix each one individually.

    Dec 2025 · comment
  • Restricts and secures agent access · 6

    MCP has a great advantage over agent using cli: MCP is much easier to secure so that it's hardwired that the agent can only call the pre-configured MCP server.

    May 2026 · comment

    Routing calls through MCPs that only allow specific access lets us keep things secure for servers that don't allow that fine control through the API.

    Oct 2026 · comment

    I've since come back to MCPs, because I want to build my own agents without first having to solve the problem of effectively sandboxing Bash.

    Sep 2026 · comment
  • Standardized OAuth and authentication · 5

    MCP in its current direction and form is really an OAuth Protocol over http. And it has something other that other agent identity protocols don't: client adoption

    May 2026 · comment

    Authorization, in my opinion, is the big problem you need MCP for, though the current MCP Authorization spec still needs some refinement.

    Mar 2026 · comment

    MCP can be just as badly designed and documented as REST, but at least there's a standard for authenticating and listing the options.

    Sep 2026 · comment
  • Easy for non-technical users · 4

    our nontechnical employees are now able to answer questions and build interactive data visualizations they would never have been able to do before

    Jul 2026 · comment

    The big advantage here is that a normal API, direct tool integration or CLI are unpractical to set up for my less technical coworkers.

    Sep 2026 · comment

    MCP can be easily updated with new tools and instructions, and it’s easy to explain how to add to Claude for non-technical people.

    Apr 2026 · comment

What people warn about

  • Consumes lots of tokens, context · 5

    Going all-in on Model Context Protocol (MCP) is definitely a bad idea because I noticed the Jira and GitLab MCPs were consuming an excessive amount of tokens.

    Sep 2026 · comment

    I find MCP beneficial too, but do be aware of token usage. With a naive implementation MCP can use significantly more input tokens

    Apr 2026 · comment

    the main problem I've seen with MCP is that they become a dumping ground and you end up with a ton of context every session

    Sep 2026 · comment
  • Clients support little beyond tools · 2

    The biggest disappointment I have with MCP today is that many clients are still half-assed on supporting the functions outside of MCP tools.

    Mar 2026 · comment

    MCP has been mostly relegated to tool calling leaving potentially powerful capabilities on the table due to lac of client support for them.

    Sep 2026 · comment
  • Security benefits questioned · 4

    how is defining access rights in an application layer the agent has access to more secure than defining the access rights at the target application itself?

    Sep 2026 · comment

    Relying on an intermediary to provide access controls, and that agents will never access the service directly, seems dangerous and naive.

    Sep 2026 · comment

    Yes, let's have the stohastic parrot guessing machine run executables on the project manager's computer - that can only end well, right?

    Dec 2025 · comment
  • Rough protocol and auth spec · 3

    Authorization, in my opinion, is the big problem you need MCP for, though the current MCP Authorization spec still needs some refinement.

    Mar 2026 · comment

    MCP's value isn't in the protocol design (which is admittedly rough). It's in what it enables: persistent, composable integrations that survive across sessions.

    Mar 2026 · comment

    MCP is simply a crappy implementation of this idea because our programming environments do not expose global remote function call mechanisms

    Jan 2026 · comment

Every quote

The 40 most detailed of 995 comments. Each links to the original.

  1. MCP in its current direction and form is really an OAuth Protocol over http. And it has something other that other agent identity protocols don't: client adoption

    Mixed · May 2026 · in “MCP Is Dead”

  2. MCP's value isn't in the protocol design (which is admittedly rough). It's in what it enables: persistent, composable integrations that survive across sessions.

    Mixed · Mar 2026 · in “MCP Is Dead; Long Live MCP”

  3. MCP has a great advantage over agent using cli: MCP is much easier to secure so that it's hardwired that the agent can only call the pre-configured MCP server.

    Recommends · May 2026 · in “MCP Is Dead”

  4. Going all-in on Model Context Protocol (MCP) is definitely a bad idea because I noticed the Jira and GitLab MCPs were consuming an excessive amount of tokens.

    Mixed · Sep 2026 · in “MCP was always a bad idea?”

  5. The whole purpose behind MCP was to provide a low-impedance standard where some set of tools could be plugged into an existing agent with no pre-knowledge

    Recommends · Nov 2025 · in “What if you don't need MCP at all?”

  6. Routing calls through MCPs that only allow specific access lets us keep things secure for servers that don't allow that fine control through the API.

    Recommends · Oct 2026 · in “Is MCP becoming the common foundation of smart, decentralised search?”

  7. our nontechnical employees are now able to answer questions and build interactive data visualizations they would never have been able to do before

    Recommends · Jul 2026 · in “AI Mania Is Eviscerating Global Decision-Making”

  8. I've since come back to MCPs, because I want to build my own agents without first having to solve the problem of effectively sandboxing Bash.

    Recommends · Sep 2026 · in “MCP was always a bad idea?”

  9. The biggest disappointment I have with MCP today is that many clients are still half-assed on supporting the functions outside of MCP tools.

    Mixed · Mar 2026 · in “MCP Is Dead; Long Live MCP”

  10. the problem isn't with MCP. As others have pointed out in this thread, you can totally have as secure an MCP server/tool as a sandboxed CLI

    Recommends · Apr 2026 · in “I Still Prefer MCP over Skills”

  11. Authorization, in my opinion, is the big problem you need MCP for, though the current MCP Authorization spec still needs some refinement.

    Mixed · Mar 2026 · in “MCP is dead. Long live the CLI”

  12. Yes, let's have the stohastic parrot guessing machine run executables on the project manager's computer - that can only end well, right?

    Mixed · Dec 2025 · in “Donating the Model Context Protocol and establishing the Agentic AI Foundation”

  13. Forcing an agent through an MCP is a way to grant access to remote services while still restricting how the network access can be used.

    Recommends · Sep 2026 · in “MCP was always a bad idea?”

  14. The big advantage here is that a normal API, direct tool integration or CLI are unpractical to set up for my less technical coworkers.

    Recommends · Sep 2026 · in “Ask HN: Who is using MCP in production?”

  15. There is huge value in having vendors standardize and simplifying their APIs instead of having agent users fix each one individually.

    Recommends · Dec 2025 · in “Donating the Model Context Protocol and establishing the Agentic AI Foundation”

  16. MCP can be just as badly designed and documented as REST, but at least there's a standard for authenticating and listing the options.

    Mixed · Sep 2026 · in “Ask HN: Who is using MCP in production?”

  17. I find MCP beneficial too, but do be aware of token usage. With a naive implementation MCP can use significantly more input tokens

    Mixed · Apr 2026 · in “Running Google Gemma 4 Locally with LM Studio's New Headless CLI and Claude Code”

  18. MCP can be easily updated with new tools and instructions, and it’s easy to explain how to add to Claude for non-technical people.

    Mixed · Apr 2026 · in “I Still Prefer MCP over Skills”

  19. I think it can be very useful to constrain what your AI can do (e.g. read files but don't delete them). MCP is a way to do that.

    Recommends · May 2026 · in “MCP Is Dead”

  20. If you have external users, then you have to use MCP. Because that's what their apps e.g. Cowork, Cursor support out-of-the-box

    Recommends · May 2026 · in “MCP Is Dead”

  21. agents don't need a protocol to connect to a server and discover its capabilities, tools, and distributed skills? I disagree.

    Mixed · Sep 2026 · in “MCP was always a bad idea?”

  22. any mcp client can connect to any mcp server without the developers on either side having to issue client ids, secrets, etc.

    Recommends · Feb 2026 · in “I made MCP cheaper in one command”

  23. I can use the same MCP servers through remote MCP on my phone, web, iPad, in ChatGPT, Perplexity, Claude, Mistral and so on

    Recommends · Apr 2026 · in “I Still Prefer MCP over Skills”

  24. the main problem I've seen with MCP is that they become a dumping ground and you end up with a ton of context every session

    Mixed · Sep 2026 · in “The Discovery Tax: Why Coding Agents Waste 2,500 Tokens Before Writing Code”

  25. its literally the only api standard that we truly made plug and play and even automatically oauth antenticathable with dcr

    Recommends · May 2026 · in “MCP Is Dead”

  26. the basic MCP approach leads to the LLM context running out of memory due to being flooded with too many tool descriptions

    Mixed · Nov 2025 · in “Claude Advanced Tool Use”

  27. That's why my OP prefaced that MCP is critical for orgs and enterprises because it alleviates some of the friction points

    Recommends · Mar 2026 · in “Show HN: Mcp2cli – One CLI for every API, 96-99% fewer tokens than native MCP”

  28. Tools are very constrained in that the agent must insert into it's context the tool call, and it can only receive the response of the tool directly back into its context.

    Criticizes · Apr 2026 · in “I Still Prefer MCP over Skills”

  29. how is defining access rights in an application layer the agent has access to more secure than defining the access rights at the target application itself?

    Criticizes · Sep 2026 · in “MCP was always a bad idea?”

  30. MCP is simply a crappy implementation of this idea because our programming environments do not expose global remote function call mechanisms

    Criticizes · Jan 2026 · in “MCP Is a Fad”

  31. MCP has been mostly relegated to tool calling leaving potentially powerful capabilities on the table due to lac of client support for them.

    Criticizes · Sep 2026 · in “You said no MCP”

  32. Relying on an intermediary to provide access controls, and that agents will never access the service directly, seems dangerous and naive.

    Criticizes · Sep 2026 · in “MCP was always a bad idea?”

  33. you are not as simple as CLI/skill/combination, and no, you are not more secure just because you are buried under 3 level of spaghetti

    Criticizes · Apr 2026 · in “I Still Prefer MCP over Skills”

  34. it has a place in its accessibility outside terminal based agents and in its ability to wrangle data before it's consumed by the agent requesting it

    Mixed · May 2026 · in “MCP Is Dead”

  35. isn't MCP effectively just a communication standard? Like the only difference between an MCP server and my Express webserver is the supposed logic

    Mixed · May 2026 · in “MCP Is Dead”

  36. The protocol provides AI a way to do this without guessing or working in different ways, because the server does all the work, deterministically.

    Recommends · Mar 2026 · in “MCP Is Dead; Long Live MCP”

  37. MCP servers are very simple to implement, and the developers of the app/service maintain the server so you don't have to create or update skills

    Recommends · Apr 2026 · in “I Still Prefer MCP over Skills”

  38. I think there's some misconception spreading across the software community that an MCP server is always inherently the optimal, obvious solution

    Mixed · Nov 2025 · in “What if you don't need MCP at all?”

  39. Agreed, MCP works and it works well. Often I'll wrap an API in an MCP because getting the agent to interact with an API just wastes tokens

    Recommends · May 2026 · in “MCP Is Dead”

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

    Mixed · Jun 2026 · in “Zero-Touch OAuth for MCP”

Counts from 1,795,640 Hacker News comments, Nov 2025 to Nov 2026. How we count · Report a mistake