這個設計來自我同時掛了五十個 MCP 工具的經驗：工具一多，Claude 挑工具就開始出錯；它從不記錄自己呼叫了什麼、為什麼；五十個工具也管不動。DB2MCP 把所有查詢收到 4 個固定工具後面：AI 用一句話說要查什麼，BM25 評分挑出該用的查詢，分數太低就說沒有、兩個太接近就反問你；call 沒寫目的就不執行，每次呼叫留一行紀錄。加再多表，對外的 4 個工具都不變，不用重新授權。 給它一條連線字串，五分鐘後 AI 就能查你的資料庫。每個答案附本機看板連結和 Excel 下載連結，都是同一支查詢再跑一次，數字可以自己核對。說「每天 8:30 給我這份」，它寫成排程報表，Claude 關著也照跑。 沒有資料庫？內建財務（應收／應付）和訂單到出貨的範例資料。有 Excel 檔？直接丟進來，每個工作表變成一張資料表。 唯讀。跑在你自己的電腦上，不用 Docker、不用 OAuth。已在 Claude Code 和 Claude Cowork 上測試，有裝 Claude Desktop 會一併註冊。
Hands-on prompt engineering and LLM evaluation. Built around a diagnose-then-ship loop: forensic transcript critique ▎ that names the exact failure mode, paired with a fixer that produces complete, deployment-ready rewrites — not ▎ vague advice.
Audits your test suite's output for silently vanished tests, unexplained skips, and suspiciously fast runs — turns green CI into a real signal.
Analyze failed CI/CD pipelines, identify likely root causes, inspect repository configuration and code context, and produce actionable, validated remediation guidance for GitHub Actions, GitLab CI, Jenkins, and related development pipelines.
Point it at your repo (or a specific change) and it scaffolds a unit-test suite for you — auto-detects your runner (Jest/Vitest, pytest, Go, XCTest, JUnit/Kotest), writes idiomatic AAA tests matching your conventions, then RUNS them and fixes test-side failures until green. It never edits your source — a real bug is reported, not papered over. Zero config.
A senior-engineer-style code review assistant: severity-ranked findings, security smell detection, and constructive feedback across 8 language families.
Academic sanity-check with interactive HTML dashboard. Detects look-ahead bias, over-parameterization, and benchmarks against 10 canonical blueprints via animated SVG gauges and color-coded KPI cards. Parses natural language, code, or CSV uploads. Output auto-adapts (EN/ZH-TW/ZH-CN). Built on De Prado (2018). For students and strategy developers. Not financial advice.