MCP Prompts -> /Slash_Commands

[ravitemer]

Hey everyone! 👋

Just like how MCP servers provide tools, resources and resourceTemplates they also provide prompts as seen below.

In addition to the server provided prompts, we can also create our own prompts using native server architecture mcphub.add_prompt api. There are two inbuilt prompts in neovim server.

A prompt in a MCP server when executed return a messages array with each message with role and content.

• role can be of two types user and assistant (Note: Native server prompts can also have system role).
• content can be of many types: text, image, audio, blob, resource

These prompts are intended to be used by the user either in slash_commands or in a menu. When the slash_command is run we should populate chat history with corresponding messages. That's exactly we are doing now.

As you can see, we can define our own conversations with these prefilled messages and on next user message the llm replicates the behaviour defined in these prefilled messages.

CodeCompanion Integration ⚡

With CodeCompanion integration, all your MCP server prompts automatically become /slash_commands just like how resources become #variables. Just add this to your config:

require("mcphub").setup({
    extensions = {
        codecompanion = {
            -- Show the mcp tool result in the chat buffer
            -- NOTE:if the result is markdown with headers, content after the headers wont be sent by codecompanion
            show_result_in_chat = true,
            make_vars = true, -- make chat #variables from MCP server resources
            make_slash_commands = true, -- make /slash_commands from MCP server prompts
        }
    }
})

Now you can:

• Use any MCP server prompt with /mcp:prompt_name
• Enter arguments through friendly vim.ui.input prompts
• Get rich responses right in your chat window
• Messages are properly formatted with correct roles (user/assistant/system)

Realtime Updates

Just like how resources automatically update as #variables, your slash commands stay in sync with your MCP servers prompts.

This creates a really smooth workflow where you can:

1. Define your prompts in your MCP servers
2. Use them directly in chats with slash commands

Resources as #variables

[bassamsdata]

Hey! First off, thanks for this amazing plugin! I was exploring it yesterday and going through the wiki, but I’m feeling a bit lost.

If you ever have the time, it would be really helpful to have a blog post or a single markdown file that walks through an Example with everything step by step, from installation to setup, usage, and even how it works with CodeCompanion. A complete workflow example would make it so much easier to understand. That would be greatly appreciated! 😊

[bassamsdata]

I think what I’m looking for might be more of a “user manual” rather than traditional documentation, similar to Notebook Setup. But since I know how challenging that can be in open source, I’ll withdraw my suggestion.

The README is quite detailed, so I really appreciate that! But when I watch the demos, I sometimes find myself unsure about why certain steps are taken and how to actually implement them. The video without narration also makes it a bit harder to follow. In the usage section, it mentions opening MCPhub and using the Hub instance API in my code, but I’m not sure exactly where that should happen.

I hope this makes sense! Your documentation is already quite thorough, and since you know the project inside and out, it might help to approach it from the perspective of someone completely new to it.

One specific suggestion: I’d strongly recommend moving the Native MCP Servers section in the wiki right after installation. That would make it clearer where things should go and why they’re needed.

[ravitemer]

documentation, similar to Notebook Setup.

@bassamsdata Thank you for the feedback. We can certainly do this! I will update the wiki as well as README giving them more time.

The video without narration also makes it a bit harder to follow

I might have to use some AI voiceover then 😅

using the Hub instance API in my code, but I’m not sure exactly where that should happen.

Sorry for the confusion. That section is supposed to be for any chat plugin author who wants to integrate with mcphub. I might have to remove that and add that in wiki in separate section.

Will try to update these as early as possible.

[bassamsdata]

We can certainly do this! I will update the wiki as well as README giving them more time.

Thank you for your understanding

I might have to use some AI voiceover then

why not, google AI studio has great voices. :)

Sorry for the confusion. That section is supposed to be for any chat plugin author

now worries! Ah okay I got it.

Thank you

[ravitemer]

@bassamsdata MCP Hub has a documentation site now finally. https://ravitemer.github.io/mcphub.nvim/

Hope this is helpful. Please check it out and let me know your thoughts.

[bassamsdata]

Congrats on the new website @ravitemer , and thank you so much for putting it together! It’s awesome, much easier to follow and understand.

[ilan-schemoul]

Seems like it doesn't work for me or I am missing a step ravitemer/mcphub.nvim#62
Maybe we should consider autointegration at this point. Usually I have noticed a huge ton of problems comes from integration between plugins in nvim. I noticed the only good pattern (where I don't waste a lot of time on integration details) is auto integration: if a plugin can detect another one he integrates with it (it can be opt in or opt out. Usually opt out is better so most users don't miss the feature)