标准的 sequential-thinking MCP 为思维链推理提供了一个很好的工具，但在多智能体环境中，它会遇到“传话游戏”问题，即智能体可能会自信地产生无根据的幻觉思维，从而污染上下文窗口。
krusch-sequential-mcp 通过引入语义合理性门控 (Semantic Plausibility Gating) 以及高度可靠的 DBOS PostgreSQL 持久化层解决了这个问题。
🧠 语义合理性门控： 通过边缘模型评估器自动拒绝偏离或产生幻觉的思维。
💾 DBOS PostgreSQL 持久化： 将每一个思维、分支和修订同步持久化到 dbos_thoughts 表中，创建一个可审计的推理有向无环图 (DAG)。
🛑 确定性状态可靠性： 停止被污染的思维执行，强制智能体重新评估其推理路径。
🔌 直接替换： 完全兼容标准的 sequential-thinking 接口，同时支持新的 groundingContext 参数。
📦 零外部依赖： 合理性评估器完全自包含——无需外部工具包。
Related MCP server: Tyra Advanced Memory MCP Server
当智能体提出一个思维时，内部评估器会根据提供的 groundingContext 对其进行筛选。
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];npm install -g krusch-sequential-mcp或者在您的 MCP 配置文件（例如 claude_desktop_config.json 或 .cursor/mcp.json）中进行配置：
{
"mcpServers": {
"krusch-sequential-mcp": {
"command": "npx",
"args": ["-y", "krusch-sequential-mcp"]
}
}
}智能体可以使用标准参数（thought、thoughtNumber、totalThoughts、nextThoughtNeeded 等）调用 sequentialthinking 工具。
要启用合理性门控，请在工具调用中包含 groundingContext 参数：
{
"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."
}由于该思维与 groundingContext 冲突，评估器将自动拒绝它，并向智能体返回错误，要求其重新思考其方法。
变量 | 必需 | 默认值 | 描述 |
| 否 | (无 — 禁用持久化) | PostgreSQL 连接字符串（例如 |
| 否 |
| 用于合理性检查的 Ollama 服务基础 URL。 |
| 否 |
| 用于合理性筛选的 Ollama 模型。应为一个小型、快速的模型。 |
复制 .env.example 以快速开始：
cp .env.example .env我们欢迎贡献！请确保您的测试通过并遵守项目格式标准。
通过 npm run build 和 npm start（或 node build/index.js）运行测试。
MIT License © 2026 kruschdev
A detailed tool for dynamic and reflective problem-solving through thoughts. Augmented with Semantic Plausibility Gating.
| Name | Required | Description | Default |
|---|---|---|---|
| thought | Yes | Your current thinking step | |
| branchId | No | Branch identifier | |
| isRevision | No | Whether this revises previous thinking | |
| thoughtNumber | Yes | Current thought number | |
| totalThoughts | Yes | Estimated total thoughts needed | |
| revisesThought | No | Which thought is being reconsidered | |
| groundingContext | No | OPTIONAL: Provide the source context for this thought. The server will independently verify the plausibility of your thought against this context. | |
| branchFromThought | No | Branching point thought number | |
| needsMoreThoughts | No | If more thoughts are needed | |
| nextThoughtNeeded | Yes | Whether another thought step is needed |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations exist, so the description carries the full burden. It mentions 'Semantic Plausibility Gating' which hints at a verification mechanism, but does not explain how it works, what data is stored, or any side effects. The schema's groundingContext field description provides some behavior, but the main description is insufficient for an agent to understand the tool's operational characteristics.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is very short at one sentence, which is lean but not necessarily well-structured. It front-loads the core concept, but the single sentence lacks detail that could be organized in a more informative way. It is not overly verbose, but it doesn't make effective use of its brevity.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a complex tool with 10 parameters and no output schema, the description is inadequate. It doesn't explain the workflow (e.g., how to sequence thoughts, the meaning of branchId, revision, nextThoughtNeeded), nor does it describe the plausibility gating behavior beyond naming it. The schema helps but the description leaves the agent without context on how to use the tool effectively.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
All 10 parameters have schema descriptions, so the baseline is 3. The main description adds no parameter-specific meaning, and it doesn't mention any relationships between params. However, given 100% schema coverage, the schema itself is sufficient for understanding parameters.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description describes the tool as 'for dynamic and reflective problem-solving through thoughts', which conveys the general domain but lacks a specific action verb (e.g., submit, record) or resource. The name 'sequentialthinking' partially compensates, but the description alone doesn't clarify what the tool does beyond generic problem-solving.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance on when to use this tool versus other tools. No alternatives are mentioned, and no context is provided about the appropriate use case. The description simply states what it is without indications of when it should be preferred.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v1.0.0sequentialthinkingScored across 1 tool
Only one tool exists, so there is no possibility of confusion or overlap. The tool's purpose is singular and clear.
With a single tool, naming consistency is trivially satisfied. The name 'sequentialthinking' is descriptive and matches the server's focus.
The server has exactly one tool, which feels minimal. While it is appropriate for a focused sequential thinking utility, the count is on the borderline of being too thin.
The tool appears to provide a comprehensive capability for sequential thinking and problem-solving. However, being the only tool, there may be missing auxiliary operations like reset or history, though no obvious gaps are evident.
Cloud-hosted MCP server for durable AI memory
Hosted MCP server connecting claude.ai, ChatGPT and other AI apps to your own computer
MCP server for AI dialogue using various LLM models via AceDataCloud
Capability registry for the agentic economy. Semantic search over verified MCP server listings.
