The xingce-solver server is an assistive toolkit for LLM agents to tackle Chinese civil service exam (行测) questions. It provides structured routing, method retrieval, solving scaffolds, and constrained prompt generation—without directly answering questions itself.

• Question Routing (route_xingce_question): Classify a question into the appropriate exam module (e.g., 资料分析, 逻辑判断, 图形推理) with confidence scores and recommended analysis tracks; supports module_hint and section_context for guided routing.

• Lightweight Classification (classify_question): Quickly route a question stem to preliminary modules and relevant method cards via keyword matching.

• Method Card Retrieval (get_method_card, search_methods): Fetch full method cards by ID or search the 292-card knowledge base by keyword, with optional module filtering.

• Source Reference Lookup (get_source_reference): Retrieve source files, page references, confidence levels, and review flags for a given method.

• Structured Solving Drafts (solve_data_analysis, solve_logic_reasoning): Build step-by-step solving drafts for 资料分析 and 逻辑判断 questions that an LLM can follow to produce a final answer.

• Methodological Scaffolds: Read-only step-by-step guides for six modules—Graphic Reasoning, Definition Judgement, Analogy Reasoning, Logic Analysis, Quantity Relation, and Verbal Reasoning—covering observation order, checklists, response templates, and uncertainty policies.

• Prompt Composition (compose_xingce_analysis_prompt, compose_xingce_answer_prompt): Generate structured analysis or conservative answer prompts incorporating routing results, scaffold summaries, strict output schemas, and safety contracts.

• Safety Gates: Hard-coded checks block answering when visual content or materials are missing, routing is uncertain, or answer mode is disabled—returning a block reason instead of a guess.

1. Click on "Install Server".

2. Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.

3. In the chat, type @ followed by the MCP server name and your instructions, e.g., "@xingce-solver 资料分析：2020年收入132亿元，同比增10%，求2019年收入？"

That's it! The server will respond to your query, and you can continue using it as needed.

Here is a step-by-step guide with screenshots.