MCP 服务器

krusch-sequential-mcp

An advanced MCP server for reflective chain-of-thought with Semantic Plausibility Gating and DBOS PostgreSQL persistence.

README

Krusch Sequential MCP

An advanced Model Context Protocol (MCP) server for reflective chain-of-thought, augmented with Semantic Plausibility Gating and DBOS PostgreSQL persistence.

⚡ Why Krusch Sequential MCP?

The standard sequential-thinking MCP provides a great tool for chain-of-thought reasoning, but it suffers from the "Telephone Game" problem in multi-agent environments, where agents can confidently hallucinate ungrounded thoughts that poison the context window.

krusch-sequential-mcp solves this by introducing Semantic Plausibility Gating alongside a highly reliable DBOS PostgreSQL persistence layer.

Key Features

🧠 Semantic Plausibility Gating: Autonomously rejects drifted or hallucinated thoughts via an edge model evaluator.
💾 DBOS PostgreSQL Persistence: Synchronously persists every thought, branch, and revision into a dbos_thoughts table, creating an auditable DAG of reasoning.
🛑 Deterministic State Reliability: Halts poisoned thought execution, forcing agents to re-evaluate their reasoning path.
🔌 Drop-In Replacement: Fully compatible with the standard sequential-thinking interface while supporting the new groundingContext parameter.
📦 Zero External Dependencies: The plausibility evaluator is fully self-contained — no external toolkit required.

🧠 Architecture: Semantic Plausibility Gate

When an agent proposes a thought, the internal evaluator screens it against the provided groundingContext.

graph TD;
    A[Agent Thought Proposed] --> B{Grounding Context Provided?};
    B -- No --> C[Accept & Persist to DBOS];
    B -- Yes --> D[Edge Model Evaluator];
    D -- Plausible --> C;
    D -- Hallucinated/Drifted --> E[Reject Thought];
    E --> F[Return Soft Error to Agent];
    F --> G[Agent Re-evaluates];

📦 Installation

npm install -g krusch-sequential-mcp

Or configure it in your MCP settings file (e.g., claude_desktop_config.json or .cursor/mcp.json):

{
  "mcpServers": {
    "krusch-sequential-mcp": {
      "command": "npx",
      "args": ["-y", "krusch-sequential-mcp"]
    }
  }
}

🚀 Quick Start Guide

Agents can invoke the sequentialthinking tool with the standard parameters (thought, thoughtNumber, totalThoughts, nextThoughtNeeded, etc.).

To engage the plausibility gate, include the groundingContext parameter in your tool call:

{
  "thought": "Since the user is asking about the database schema, I will assume it uses MongoDB and write a query for it.",
  "thoughtNumber": 1,
  "totalThoughts": 3,
  "nextThoughtNeeded": true,
  "groundingContext": "The current codebase exclusively uses DBOS PostgreSQL for persistence. No NoSQL databases are present."
}

Because the thought conflicts with the groundingContext, the evaluator will autonomously reject it, returning an error to the agent to rethink its approach.

⚙️ Environment Variables

Variable	Required	Default	Description
`DATABASE_URL`	No	(none — persistence disabled)	PostgreSQL connection string (e.g., `postgres://user:pass@localhost:5432/mydb`). If not set, the server runs in memory-only mode.
`OLLAMA_URL`	No	`http://localhost:11434`	Base URL to the Ollama service for plausibility checks.
`PLAUSIBILITY_MODEL`	No	`qwen2.5-coder:1.5b`	Ollama model used for plausibility screening. Should be a small, fast model.

Copy .env.example for a quick start:

cp .env.example .env

🤝 Contributing

We welcome contributions! Please ensure your tests pass and adhere to the project formatting standards. Run tests via npm run build and npm start (or node build/index.js).