Advanced story classification MCP server using FastMCP, intelligent scraping, automated source discovery, and human-in-the-loop learning
This server has significant quality gaps across naming, descriptions, schema completeness, and error handling. While tool names generally follow verb patterns (discover_stories, scrape_story, classify_story), the definitions lack rigor expected of production tools. Parameter descriptions are often vague or missing type constraints. Output schemas are not documented. Error handling provides no recovery guidance. The server implements 8 tools with moderate naming clarity but poor schema coverage and minimal LLM-optimized documentation.
Classify a story into analytical frameworks and return confidence scores
End-to-end workflow: discover, scrape, classify, and save stories on a topic
Detect data visualizations in HTML by inspecting img tags and figcaption elements
Search for stories on a given topic using Google Custom Search API with diverse queries and quality filtering
Evaluate if content is a data story based on visualization count and data terminology
Record human feedback on story classification for model training
Save classified stories and their classifications to an Excel workbook
Output schemas not documented for any tool. LLMs cannot plan downstream calls without knowing what fields and types to expect in responses.
Parameter descriptions lack type constraints (ranges, enums, formats). E.g. max_results in discover_stories has no upper bound; frameworks in classify_story are undefined; filename in save_stories_excel lacks guidance on overwrite behavior.
Tool descriptions are generic and lack LLM-optimized guidance. Most are 45-65 characters, below the baseline of 194 chars for A-grade tools. Descriptions do not answer WHEN to use a tool or what problem it solves. E.g. scrape_story vs classify_story distinction is unclear.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 43 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Scrape a story from a given URL and extract title, content, metadata, and detect data elements
No error handling or recovery guidance. Tools do not specify what errors are possible, whether they are retryable, or what the LLM should do next. E.g. scrape_story can fail on unreachable URLs, malformed HTML, or API throttling, none documented.
Idempotency not documented for write operations. Tools like save_stories_excel and provide_feedback modify state but do not clarify behavior on retry (overwrite? append? error?).
Complex input objects (Story, Classification) are referenced but not documented. LLMs do not know which fields are required, types, or constraints (e.g. what is a valid framework name?).
Tool composition issues. collect_stories is a wrapper around discover_stories → scrape_story → classify_story → save_stories_excel. While composition is valid, intermediate tools lack clear delegation hints. LLMs may not know whether to call collect_stories (one-shot) or chain tools (for fine control).
Secrets (Google API key, OpenAI key) are injected via environment variables, which is correct. However, no documentation of secret names or how to configure them.