MCP Tutorial (No Command Line)
This guide is for beginners.
The goal is simple:
- no terminal
- no commands
- only the NextClaw UI
You will complete one full MCP flow:
- install
- check
- use it in chat
If you only want the short version, remember this:
- open
Marketplace -> MCP - click
Install - run
Doctorbefore trying to use it in chat
1) What MCP means here
MCP is a standard way to give your AI external tool capabilities.
It is not a separate chat mode.
The real flow is:
- install an MCP server
- verify it is reachable
- let the agent use it as part of its toolset
So the mental model should be:
- MCP is a tool source
- chat is still where you actually use it
2) Open the MCP marketplace
In the left sidebar of NextClaw:
- open
Marketplace - then open
MCP
You will usually see two areas:
MCP Market: installable MCP itemsInstalled MCP: MCP servers you already installed
Each card usually shows:
- name
- short summary
- transport type such as
STDIO,HTTP, orSSE - current status
3) Install one MCP
Find an item in the MCP market and click Install.
The install dialog usually includes:
Server NameAvailable to All Agents- optional input fields required by that MCP
For beginners, the safest default is:
- keep the default server name
- keep
Available to All Agentsenabled - only change extra inputs if you clearly know what they do
After install, the item should appear in:
Installed MCP
In most cases, you do not need to restart the service.
4) Always run Doctor first
Installed does not always mean usable.
The most important next step is:
- click
Doctoron the installed MCP card
Pay attention to:
AccessibleTransportTools
The ideal result is:
Accessible = trueTools > 0
That means:
- NextClaw can reach the MCP server
- and it can already discover tools from it
If Doctor fails, do not assume chat will still use it correctly.
5) How to actually use it after that
Once Doctor passes, go back to chat.
Important:
- you do not open a separate “MCP chat”
- you do not switch into a special MCP mode
You just talk to the agent normally.
Examples:
- “Use the browser tools to inspect this page.”
- “Check this external service for me.”
- “Use the installed tool to verify the current state.”
If your current runtime/agent supports MCP, the installed MCP server becomes an extra tool source.
6) Chrome DevTools MCP as an example
Chrome DevTools MCP is a good example of how this works.
You can install it from the MCP marketplace without touching the command line.
But there is one important detail:
- some MCP servers depend on external programs or system state
For Chrome DevTools MCP, that usually means Chrome itself must be in a connectable state.
So the correct flow is:
- install it in UI
- run
Doctor - if
Doctorfails, fix the browser-side requirement first
So for beginner users:
- installation can be no-command-line
- usability still depends on the MCP's own prerequisites
Doctoris the final truth
7) What Enable / Disable / Remove mean
After install, you will usually see actions like:
DisableDoctorRemove
Think of them like this:
Disable- temporarily stop using it
- keep the config
- re-enable later
Doctor- test whether it is actually reachable and usable
Remove- delete the MCP config entirely
- use this if you installed the wrong one or want to start over
If you only want to pause it, use Disable.
If you are done with it, use Remove.
8) Beginner-friendly order
For your first MCP setup, use this order:
- choose an official or well-documented MCP item
- install with default values
- run
Doctorimmediately - only test in chat after
Doctorpasses - if it fails, check the MCP's external prerequisite before blaming chat
9) Best next step
If this is your first time using MCP, do this:
- open
Marketplace -> MCP - choose the simplest item you understand
- click
Install - click
Doctor - if it becomes accessible, go to chat and test it once