Sequential Thinking
An MCP server designed for sequential thinking and problem-solving tasks.
Sequential Thinking MCP Server - Streamable HTTP Implementation
An MCP server that provides a powerful sequential thinking tool for complex problem-solving through thoughts.
This server enables AI assistants to work through problems step-by-step, with the ability to revise, branch, and explore alternative approaches as understanding deepens.
Features
- Dynamic Thought Process: Adjust the number of thoughts as you progress
- Revision Capability: Question and revise previous thoughts
- Branching Logic: Explore alternative approaches
- Flexible Planning: Add more thoughts even after reaching initial estimates
- Context Preservation: Maintain context across multiple thinking steps
Getting Started
To get started with this server, clone the repository and install the necessary dependencies.
git clone https://github.com/camilovelezr/server-sequential-thinking.git
cd server-sequential-thinking
npm install
Usage
Streamable HTTP Transport
The server uses the modern Streamable HTTP transport protocol for communication. Start the server:
npm start
The server will start on http://127.0.0.1:3000 by default. You can change the port using the PORT environment variable:
PORT=8080 npm start
Endpoints
- MCP Endpoint:
POST/GET/DELETE http://127.0.0.1:3000/mcp- Main MCP communication endpointPOST- Send messages to the server (initialize, tool calls, etc.)GET- Establish SSE stream for server-to-client notificationsDELETE- Terminate session
- Health Check:
GET http://127.0.0.1:3000/health- Check server status and active sessions - Test Interface:
GET http://127.0.0.1:3000/- Browser-based test interface
Transport Features
The Streamable HTTP transport provides several advantages:
- Session-Based State: Manages state for each client session in memory. The core server is long-running and can be scaled with sticky sessions.
- Flexible Streaming: Server can upgrade any response to SSE for streaming
- Infrastructure Compatible: Works with standard HTTP middleware and proxies
- Resumable Connections: Support for connection resumption
- Automatic Session Cleanup: Inactive sessions and their associated data are automatically cleared after 1 hour to conserve resources.
Security
The implementation includes security measures:
- Origin header validation to prevent DNS rebinding attacks
- Server binds only to localhost (127.0.0.1)
- CORS headers are properly configured
- Session IDs are cryptographically secure
- Privacy by Design: No sensitive tool inputs or outputs are logged to the console.
Adding to Cursor
To use this server with Cursor, add the following to your Cursor configuration:
{
"mcpServers": {
"sequential-thinking": {
"url": "http://localhost:3000/mcp"
},
}
}
Running with Docker
You can also run the server in a containerized environment using Docker for easier deployment and isolation.
-
Build the Docker image:
docker build -t mcp-server-sequential-thinking . -
Run the Docker container:
docker run -p 3000:3000 mcp-server-sequential-thinking
The server will be accessible at http://127.0.0.1:3000.
Available Tool
sequentialthinking
A detailed tool for dynamic and reflective problem-solving through thoughts.
When to use:
- Breaking down complex problems into steps
- Planning and design with room for revision
- Analysis that might need course correction
- Problems where the full scope might not be clear initially
- Multi-step solutions requiring context preservation
Parameters:
thought(string, required): Your current thinking stepnextThoughtNeeded(boolean, required): Whether another thought step is neededthoughtNumber(integer, required): Current thought numbertotalThoughts(integer, required): Estimated total thoughts neededisRevision(boolean, optional): Whether this revises previous thinkingrevisesThought(integer, optional): Which thought is being reconsideredbranchFromThought(integer, optional): Branching point thought numberbranchId(string, optional): Branch identifierneedsMoreThoughts(boolean, optional): If more thoughts are needed
License
MIT
関連サーバー
Kone.vc
スポンサーMonetize your AI agent with contextual product recommendations
CalmBoard.ai
MCP - Multi-purpose endpoint for AI agents to access CalmBoard's features and data. This MCP provides a comprehensive interface for AI agents to interact with CalmBoard's project/scrum/kanban board management functionalities and retrieve relevant data to provide insights and recommendations to users.
cli-anything-zotero
CLI & MCP server for Zotero 7/8 — 52 tools to let AI manage your research library locally. Search, import, export, PDF, notes, and more.
YNAB (You Need A Budget)
An MCP server for YNAB (https://www.ynab.com/), exposing 20 tools for querying and managing your budget through any MCP-compatible client (Claude Desktop, etc.).
Wise MCP Server
A gateway for the Wise API to manage recipients, requiring a Wise API token.
Confluence
Integrate with Atlassian Confluence to access spaces, search pages, and manage content from any MCP-compatible application.
screenpipe
use 24/7 desktop memory as context in AI
iTerm MCP
Provides access to your iTerm session, requiring iTerm2 and Node.js.
ReaperMCP
AI-powered music production in REAPER via the Model Context Protocol — 150 tools for composition, mixing, mastering, and audio analysis.
Atlassian
Interact with Atlassian tools like Confluence and Jira.
Time Server
Get the current time and convert time between different timezones.