MCP server · Reference servers

Sequential Thinking MCP server

Gives an assistant a tool for step-by-step reasoning that it can revise or branch as it works through a problem.

Maintainer: Model Context Protocol project (official)
Licence: Apache-2.0 AND MIT
Runs: Locally (npx)
Sign-in: None
Last activity: 2026-10-11
Source: github.com/modelcontextprotocol/servers/tree/main/src/sequentialthinking

Overview

What it does

Sequential Thinking is a Node.js reference server from the MCP project with a single tool, sequential_thinking. The model calls it once per step, sending the current thought, its number, an estimate of how many steps remain and whether another is needed. It can mark a step as a revision of an earlier one, branch from a previous step under a branch ID, or raise the total when the problem turns out larger than expected.

It helps with planning, design and debugging tasks where the full scope isn’t clear at the start and the plan may need correcting. You don’t call the tool yourself; you ask the assistant to think through a problem step by step and the host decides when to use it. The README suggests checking your client’s tool log for repeated calls to confirm it is working. Set DISABLE_THOUGHT_LOGGING to true to stop the server logging each thought.

Tools

What the assistant can do

Record numbered reasoning steps with an adjustable total
Revise an earlier step
Branch into alternative lines of reasoning
Signal when more steps are needed

Install

Install Sequential Thinking MCP server in your app

You need:

Node.js with npx

Run in your project. --scope local keeps the servers private to you in this project; use user for every project.

Terminal

claude mcp add --scope local sequential-thinking -- npx -y @modelcontextprotocol/server-sequential-thinking

Native Windows (PowerShell)

claude mcp add --scope local sequential-thinking -- cmd /c npx -y '@modelcontextprotocol/server-sequential-thinking'

Project root. Commit it to share the servers with your team; each person sets the variables in their own shell.

.mcp.json

{
  "mcpServers": {
    "sequential-thinking": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-sequential-thinking"
      ]
    }
  }
}

On native Windows (not WSL), the reference servers’ READMEs start npx through `cmd /c`: if a server won’t start, use "command": "cmd" with "args": ["/c", "npx", …].

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json (Settings → Developer → Edit Config). Restart Claude after saving.

claude_desktop_config.json

{
  "mcpServers": {
    "sequential-thinking": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-sequential-thinking"
      ]
    }
  }
}

In your project, or ~/.cursor/mcp.json for every project.

mcp.json

{
  "mcpServers": {
    "sequential-thinking": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-sequential-thinking"
      ]
    }
  }
}

In your workspace. VS Code asks for each secret the first time a server starts and stores it securely.

mcp.json

{
  "servers": {
    "sequential-thinking": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-sequential-thinking"
      ]
    }
  }
}

macOS/Linux: ~/.config/devin/mcp_config.json · Windows: %APPDATA%\devin\mcp_config.json. Read by the default Devin Local agent and by the legacy Cascade agent.

mcp_config.json

{
  "mcpServers": {
    "sequential-thinking": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-sequential-thinking"
      ]
    }
  }
}

Want it alongside other servers? Open Sequential Thinking MCP Server in the MCP config generator and add the rest; it writes one file per app. The snippets follow the setup in the server’s README, checked on 2026-10-14.

Before you install

It has no file, network or shell access; it only records the thoughts the model sends.
Thoughts are logged by the server by default; set DISABLE_THOUGHT_LOGGING to true if they may contain sensitive details.
A local MCP server runs with your user account’s permissions. Install it only from the source linked above. See how to choose a safe server.

Why it’s listed

Why we list it

A safe, dependency-free way to encourage structured multi-step reasoning, kept by the MCP project.

All MCP servers MCP config generator Guide: MCP servers in Claude Code Guide: what is MCP?

What it does

What the assistant can do

Install Sequential Thinking MCP server in your app

Before you install

Why we list it

More MCP servers