Sequential Thinking MCP ServerFeatured

What is Sequential Thinking MCP Server?

The Sequential Thinking MCP server is an official reference server from the Model Context Protocol project that gives a model one tool for structured, step-by-step problem solving, with room to revise earlier steps and branch into alternative lines of reasoning.

Instead of answering in a single pass, the model records numbered thoughts, adjusts how many it expects to need, marks revisions and can explore branches before it commits to an answer. It is published on npm as @modelcontextprotocol/server-sequential-thinking and as the Docker image mcp/sequentialthinking. It needs no API key and calls no external service.

How to use Sequential Thinking MCP Server?

Prerequisites: Node.js (for npx) or Docker. There are no environment variables to set, apart from an optional DISABLE_THOUGHT_LOGGING=true to stop the server from logging thought content.

Claude Desktop

Add to claude_desktop_config.json:

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

On Windows, launch npx through cmd:

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

With Docker, use "command": "docker" and "args": ["run", "--rm", "-i", "mcp/sequentialthinking"].

Claude Code

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

Cursor

Cursor uses the same mcpServers format. Add the npx block above to ~/.cursor/mcp.json (global) or .cursor/mcp.json (project).

VS Code

Run MCP: Open User Configuration from the Command Palette, or create .vscode/mcp.json in your workspace:

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

Codex

codex mcp add sequential-thinking npx -y @modelcontextprotocol/server-sequential-thinking

After adding the server, restart or reload your client, confirm sequential_thinking appears in its tool list, and ask for a step-by-step solution to a non-trivial problem.

Key features of Sequential Thinking MCP Server

• Breaks complex problems into numbered, manageable steps
• Lets the model revise earlier thoughts as its understanding improves
• Supports branching into alternative reasoning paths with a branch ID
• Grows or shrinks the planned number of thoughts while working
• Encourages generating a solution hypothesis and then verifying it
• Runs entirely locally with no API key or external service
• Available via npx and Docker; MIT-licensed

Sequential Thinking MCP Server tools

The server exposes a single tool, sequential_thinking. Each call carries one step of reasoning:

• Core fields: thought (the current step), thoughtNumber, totalThoughts (the current estimate) and nextThoughtNeeded (whether to keep going).
• Revision fields: isRevision and revisesThought mark a step that reconsiders an earlier one.
• Branching fields: branchFromThought and branchId start and label an alternative path.
• Extension: needsMoreThoughts signals that the model reached the planned end but needs more steps.

The host decides when to call the tool. You normally don't call it by hand unless your client exposes raw tool calls.

Use cases of Sequential Thinking MCP Server

• Migration planning: "Plan a database migration from PostgreSQL 14 to 16, list risks, and revise the plan if downtime exceeds 5 minutes."
• Debugging: "Debug why this deployment only fails in production and show your reasoning step by step."
• Architecture decisions: "Compare three architecture options for a file sync engine and branch if one assumption turns out to be wrong."
• Planning and design work where the scope is unclear at the start and the plan needs room for course correction.
• Long multi-step tasks that need to keep context across steps and filter out irrelevant information.

Sequential Thinking MCP Server vs alternatives

Sequential Thinking structures reasoning inside one conversation; it does not remember anything afterwards. For persistence across sessions, pair it with the Memory MCP Server, another reference server from the same repository that stores a knowledge graph. Many models also have built-in extended reasoning modes; the sequential thinking MCP server is model-agnostic and makes each step visible as a tool call, which is useful when you want to inspect or log the reasoning. Several community forks add features such as HTTP transport, but this listing is the official reference implementation.

Troubleshooting Sequential Thinking MCP Server

• The tool doesn't appear: restart or reload the host so it reconnects, then check its MCP tool list or inspector for sequential_thinking.
• The model never calls it: ask explicitly for step-by-step reasoning on a non-trivial problem. The host decides when to use the tool.
• npx fails to start on Windows: use "command": "cmd" with "/c", "npx", ... as shown above.
• Thought text in logs: set DISABLE_THOUGHT_LOGGING to true to stop the server from logging thought information.
• Tool name differs from docs: current releases name the tool sequential_thinking; some older listings show sequentialthinking. Update to the latest npm package or Docker image.