Sequential Thinking MCP Server

Overview

Sequential Thinking is one of the official Model Context Protocol reference servers. It exposes a single tool that structures a model's reasoning into discrete, numbered thoughts that can be revised or branched, rather than forcing a single-shot response to a complex problem.

How it works

1. The client calls the sequential_thinking tool once per reasoning step, passing the current thought, its thoughtNumber, an estimated totalThoughts, and whether nextThoughtNeeded is true.
2. A thought can mark itself as a revision of an earlier one (isRevision, revisesThought) or branch off an earlier thought (branchFromThought, branchId), so reasoning can backtrack or explore alternatives.
3. The host decides how many times to call the tool while it works through a problem; you do not call it by hand unless your client exposes raw tool calls.
4. Fits problems like planning a migration with risk analysis, debugging an environment-specific failure, or comparing architecture options where an early assumption might turn out wrong.

Examples

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

Installation

You need Node.js 18+.

1. Add the server to your MCP client's config using the block in the sidebar.
2. Restart your client.
3. There's nothing else to configure: the client calls the sequential_thinking tool automatically when it's working through a complex problem.