mcp-sequentialthinking-tools
@spences10
🧠An adaptation of the MCP Sequential Thinking Server to guide tool usage. This server provides recommendations for which MCP tools would be most effective at each stage.
Overview
What is mcp-sequentialthinking-tools?
A lightweight MCP server for recording sequential reasoning steps. It functions as a scratchpad with history, branching, revision metadata, and optional validation for model-authored tool plans. It does not discover other MCP tools or choose tools for the model.
How to use mcp-sequentialthinking-tools?
Install and configure via npx -y mcp-sequentialthinking-tools in your MCP client settings (e.g., Claude Desktop). Optionally set the MAX_HISTORY_SIZE environment variable (default 1000 per session). Use the provided sequentialthinking_tools, get_thinking_history, and clear_thinking_history tools, and optionally include the sequential-thinking-guidance prompt.
Key features of mcp-sequentialthinking-tools
- Records sequential thoughts with step numbering and estimates.
- Supports branching and revision of reasoning steps.
- Validates model‑recommended tool names against a supplied list.
- Sessions isolate history; defaults to
"default". - Security filtering redacts prompt‑injection‑like text.
- History retrieval and clearing via dedicated tools.
Use cases of mcp-sequentialthinking-tools
- Breaking a messy problem into inspectable reasoning steps.
- Revising or branching a plan during complex tasks.
- Keeping a small reasoning history by session.
- Validating tool‑plan names against a known tool list.
- Clearing or inspecting reasoning history during a long agent run.
FAQ from mcp-sequentialthinking-tools
How does this server differ from tools that auto‑select tools?
It does not discover or choose tools for the model. It only records thoughts and validates that recommended tool names exist in the supplied available_tools list.
What runtime dependencies are required?
The server runs on Node.js and is invoked via npx. It uses the tmcp library for MCP transport.
Where is the thinking history stored?
History is kept in memory per session. The maximum number of entries per session is configurable via the MAX_HISTORY_SIZE environment variable (default 1000).
What security measures are included?
Thought text, tool descriptions, rationales, and remaining‑step text are scanned for prompt‑injection‑like content. Matching fields are redacted before storage or retrieval, and calls with redactions include security_warnings.
What transports does the server support?
It uses a stdio transport that accepts both standard Content‑Length framed MCP messages and newline‑delimited JSON used by older tmcp tooling.
