• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    MCP versus command-line interfaces for AI integrations

    A user argues Model Context Protocol (MCP) tools are better for most integrations, citing improved model tool calling, the ability to defer tools and MCP now being stateless.

    tobi lutkeTL
    elvisEL
    ThariqTH
    9 Sources, ,

    TLDR

    A user makes the case for MCPs over command-line interfaces for most integrations—a preference they say they didn't expect. They credit better model tool calling, the ability to defer tools and MCP now being stateless. For composing or filtering data, they recommend adding parameters such as “query” to MCP tools.

    Combined views

    860.5K

    9 Sources, first seen 22d ago

    Combined views

    860.5K

    9 Sources, first seen 22d ago

    4.2K likes
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    22d ago
    first seen 22d ago
    4.2K likes
    445 comments
    1.9K saves
    187 reposts
    445 comments
    1.9K saves
    187 reposts

    9 Sources

    Thariq@trq212I was not expecting things to go this way, but I think MCPs are better than CLIs for most integrations. The models have gotten much better at tool calling, we can defer tools & MCP is now stateless. If you need to compose/filter data, add params like query to your MCP tools.22d
    elvis@omarsar0I agree. MCP is clearly better than CLI for most integrations. It felt that way early on, and it still does now. And frontier models keep getting better at it. If you build custom harnesses, use MCP tools. My meta harnesses use MCP for all external tools.22d
    tobi lutke@tobi@trq212 i think this is true as long as the models can use the mcp(s) through some form of repl. Without that, bash and cli can act as a sort of repl.22d
    Machine Learning Street Talk@MLStreetTalkDisagree. It's all about agentic CLIs. Sorry 😃 More robust, more sophisticated, fewer hallucinations, better in every imaginable way as far as I'm concerned. I can understand from Anthropic's point of view wanting to build a serverless/SaaS platform that MCP serves their purposes, but it doesn't serve ours. You have to do a bit of infra work to work with CLIs but it pays off dramatically.22d
    Jason ✨👾SaaStr.Ai✨ Lemkin@jasonlkControversial but … a human “MCPing” into many B2B apps isn’t worth it other than for a quick check / answer: - Slow - Can only “see” subset of app, if really anything at all Much more valuable: Having your agent MCP in and just figure it out and do it for you22d

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    9 Sources

    Thariq@trq212I was not expecting things to go this way, but I think MCPs are better than CLIs for most integrations. The models have gotten much better at tool calling, we can defer tools & MCP is now stateless. If you need to compose/filter data, add params like query to your MCP tools.22d
    elvis@omarsar0I agree. MCP is clearly better than CLI for most integrations. It felt that way early on, and it still does now. And frontier models keep getting better at it. If you build custom harnesses, use MCP tools. My meta harnesses use MCP for all external tools.22d
    tobi lutke@tobi@trq212 i think this is true as long as the models can use the mcp(s) through some form of repl. Without that, bash and cli can act as a sort of repl.22d
    Machine Learning Street Talk@MLStreetTalkDisagree. It's all about agentic CLIs. Sorry 😃 More robust, more sophisticated, fewer hallucinations, better in every imaginable way as far as I'm concerned. I can understand from Anthropic's point of view wanting to build a serverless/SaaS platform that MCP serves their purposes, but it doesn't serve ours. You have to do a bit of infra work to work with CLIs but it pays off dramatically.22d
    Jason ✨👾SaaStr.Ai✨ Lemkin@jasonlkControversial but … a human “MCPing” into many B2B apps isn’t worth it other than for a quick check / answer: - Slow - Can only “see” subset of app, if really anything at all Much more valuable: Having your agent MCP in and just figure it out and do it for you22d