Sequential Thinking MCP Server

How to use

Use the server with an MCP client to record your thoughts in a defined sequence and to build structured summaries. You will log each thought as you progress through the stages, review how thoughts relate to each other, and then generate a compact overview of the entire thinking process.

Primary actions you will perform: - Record a new thought with its stage, position in the sequence, and any tags or axioms used. - Check progress as you move through Problem Definition, Research, Analysis, Synthesis, and Conclusion. - Generate a summary to capture the full thinking trajectory. - Clear the history if you want to reset the session and start fresh.

Typical workflow: 1) Start a new thinking session and begin with the first thought at the Problem Definition stage. 2) Add subsequent thoughts in order, marking whether more thoughts are needed after each one. 3) Review the connections between thoughts using the analysis features. 4) When ready, generate a concise summary that covers all stages and thoughts.

You can access the main tools through your MCP client: - process_thought to record and analyze a new thought - generate_summary to produce an overview of the entire thinking process - clear_history to reset the session when needed

How to install

Prerequisites you need before starting: - Python 3.10 or higher - UV package manager (for installing and running the server)

Install and set up the project locally by following these steps in order:

uv venv
.venv\Scripts\activate  # Windows
source .venv/bin/activate  # Unix-like

uv pip install -e .

# For development with testing tools
uv pip install -e ".[dev]"

# For all optional dependencies
uv pip install -e ".[all]"

Configuration and usage options

Start the standard MCP server directly in your environment using the run interface for local I/O. You can also run a dedicated SSE HTTP transport variant if you prefer HTTP-based communication.

# Standard mode (stdio) using the built-in server module
uv run -m mcp_sequential_thinking.server

# Or use the installed CLI script
mcp-sequential-thinking

To run the HTTP Transport (SSE) variant, start the server with the transport option and a port number, then connect to it via HTTP for SSE streaming.

uv run -m mcp_sequential_thinking.server --transport sse --port 8000

# Or use the dedicated SSE script
python run_sse_server.py --port 8000

# Test SSE connection
python test_sse_connection.py --url http://localhost:8000

Additional notes

The server automatically handles data persistence, provides progress tracking, and supports importing/exporting thinking sessions. It validates thoughts with structured data models and offers tools to analyze relationships between ideas.

You can also run tests to verify functionality during development, such as the analysis and storage components, and run coverage reports to gauge test completeness.

Troubleshooting tips

If you encounter connection issues, verify that the server process is running and that the port you are using is available. Check the transport mode you selected and ensure your MCP client is configured to communicate with the correct endpoint.

If data persistence fails, ensure you have write permissions to the storage path and that any thread-safety locks are functioning as expected in your operating system. Review logs for errors related to validation or storage access.

Security considerations

Limit access to the server to trusted clients, especially when using the HTTP transport. Use secure networks and consider implementing authentication if your environment requires multi-user use.

process_thought

Records and analyzes a new thought in your sequential thinking process, including its stage, position, and optional metadata like tags and axioms used.

generate_summary

Generates a concise summary of the entire thinking process, including stage distribution and a chronological timeline.

clear_history

Resets the thinking session by clearing all recorded thoughts.