Loading ToolBench

Reading the current corpus…

mcp-chat-insight MCP server · F · ToolBench

mcp-chat-insight

github.com/crazyyanchao/mcp-chat-insight

MCP server for analyzing WeChat group messages and chat data stored in MySQL databases, generating business insights and reports

2026-07-28+STDIO3 tools

32/100

Definition Quality32

Protocol Readiness28

Supportability33

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence

Current-spec patterns detected

Summary

The server defines 3 tools with mixed quality. The query tool has a basic schema but the description is in Chinese and lacks clarity on constraints. The list_tables and report tools have minimal or empty parameter schemas. Descriptions are present but lack sufficient detail for LLM comprehension. No error handling guidance, no output schemas documented, and parameter constraints are missing. The codebase shows functional logic but tool definitions do not meet production quality standards for agent-facing interfaces.

Tools (3)

list_tablesread onlysource verified30/100

获取实际使用的数据表列表。返回当前配置的实际数据表列表及其虚拟聚合表映射说明。

queryread onlysource verified42/100

对数据表执行 SQL SELECT 查询并返回结果。这些表存储的主要数据内容是指定的表描述。必须以 SELECT 开头，表名为映射表名。

reportread onlysource verified25/100

当用户需要生成群聊消息的日报、周报、月报等时，先使用该工具获取相关提示词，用于指导下一步操作。

Issues (6)

CRITICALconstrained-input ↗

This prevents LLMs from understanding parameters, constraints, or required fields.

list_tablesreport

HIGHtool ↗

Tool names are not action-verb-first. 'query' and 'report' are nouns or ambiguous gerunds. Should be 'execute_sql_query', 'run_sql_query', 'generate_report', or 'get_report_template' to signal intent and enable LLM disambiguation.

queryreport

CRITICALresponse-shaper ↗

No output/return schemas documented for any tool. LLMs cannot predict what fields to expect, cannot chain tools, and must guess how to extract and pass data downstream. E.g., query returns what? Rows as objects? Field names? Row count?

querylist_tablesreport

HIGHtool-description ↗

Descriptions are in Chinese, making them inaccessible to English-speaking LLMs and agent developers. Descriptions should be in English (or negotiated via locale). Also, descriptions lack detail: no WHEN to use, no example usage, no next steps.

Recommendations

Rename 'query' to 'execute_sql_query' to clarify action. Rename 'report' to 'get_report_guidance' or 'generate_report_template'.
Add proper input schemas for list_tables and report. Even if they have no parameters, declare {} with explicit documentation, or add optional filters (e.g., table_prefix, report_type enum).
Document output schemas for all tools. For query: return {rows: [object], row_count: int, columns: [string]}. For list_tables: return {tables: [{name: string, description: string, ddl: string}]}. For report: return {template: string, usage_guidance: string}.
Translate all descriptions to English. Restructure to follow the pattern: 'What does it do? (1 sentence) When to use? Prerequisites? Returns what?' E.g., 'Execute a MySQL SELECT query against the configured database tables. Returns rows as a list of objects with column names as keys. Call list_tables() first to discover available table mappings. Supports SELECT only; INSERT/UPDATE/DELETE are not permitted.'
Add parameter constraints and examples in schema. For query.query, add: minLength: 10, pattern: '^SELECT\s+', description: 'MySQL SELECT statement starting with SELECT, using the mapped table name (e.g., chat_db.group_messages). Will be rewritten to actual table names. DDL: [include relevant columns].'
Add error recovery hints. E.g., 'If query fails with "table not found", verify the table name matches the mapping returned by list_tables(). If connection timeout occurs, the database may be unavailable, retry after 30s or ask the user to check connectivity.'

Score history

Overall score trend

↑ 5 points across a rubric change (v1 → v2)

32/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-22	F	32	2026-07-28+	v2
2026-03-09	F	27	-	v1