Headline Vibes Analysis MCP Server

# Product Context — Headline Vibes Last updated: 2025-10-21 ## Why This Exists Decision-makers and LLM agents need a fast, standardized snapshot of market-relevant news sentiment. Raw headline feeds are noisy and inconsistent across sources. Headline Vibes provides a consistent interface that: - Normalizes and interprets sentiment (0–10 scale) - Filters for investor relevance - Surfaces political-leaning and source distribution for transparency - Supports natural language dates and monthly ranges ## Problems Solved - Fragmented sources: unify major US publishers into one analysis result - Noisy content: exclude lifestyle/irrelevant topics using relevance filters - Ambiguous sentiment: dual scoring (general vs investor-specific) - Lack of transparency: include distributions, filtering stats, and sample headlines - Friction in date selection: accept natural language (e.g., “yesterday”) ## Users - LLM agents using MCP to incorporate market sentiment - Developers integrating headline sentiment signals into analysis or dashboards - Operators reviewing sentiment trends and coverage balance across political leanings ## How It Should Work (UX/Behavior) - Discoverability: ListTools returns two tools: - analyze_headlines(input: string) - analyze_monthly_headlines(startMonth: YYYY-MM, endMonth: YYYY-MM) - Input: - analyze_headlines accepts natural language or ISO date (YYYY-MM-DD) - analyze_monthly_headlines accepts ISO months (YYYY-MM) - Output includes: - overall_sentiment.general (score + synopsis) - overall_sentiment.investor (score + synopsis + key_terms) - political_sentiments (left/center/right) with scores and counts - filtering_stats (total vs relevant, relevance_rate) - headlines_analyzed, sources_analyzed - source_distribution and political_distribution - sample_headlines_by_leaning (up to 5 per category) - Reliability: - Clear error messages for invalid dates, unparseable inputs, and EventRegistry/API failures - Conservative fallbacks: uncategorized sources default to “center” - Token budgeting + rate-limit diagnostics to respect usage caps ## Example Interactions - Daily analysis (natural language): - name: analyze_headlines, arguments: { "input": "yesterday" } - Daily analysis (specific date): - name: analyze_headlines, arguments: { "input": "2025-02-11" } - Monthly range analysis: - name: analyze_monthly_headlines, arguments: { "startMonth": "2024-01", "endMonth": "2024-12" } ## Value Proposition - Speed: one call to summarize the day’s (or month’s) investment-relevant sentiment - Clarity: normalized scales and plain-language synopses - Transparency: distributions and samples reveal coverage patterns - Consistency: deterministic relevance filters and lexicon-based investor score ## Usage Guidance - For snapshots of a single day’s headlines, use analyze_headlines - For trend-like aggregations across months, use analyze_monthly_headlines - For cost control, review token diagnostics before triggering large ranges. - Ensure NEWS_API_KEY (EventRegistry) is configured in MCP settings. ## Out of Scope (Product) - Topic/entity-level breakdown (per ticker or sector) - Visualization/graphing (external responsibility) - Persistent storage or analytics UI (this Memory Bank is documentation-only) - Real-time streaming or long-polling ## KPIs / Signals of Success - Low error rate for natural language parsing and API requests - Stable relevance_rate across typical days (indicates balanced filtering) - Coverage balance across political leanings (non-zero representation when available) - Reasonable headlines_analyzed and sources_analyzed counts without rate-limit errors ## Future Opportunities - Caching/memoization of results for specific dates/months - Time-series trend comparisons and change explanations - Expanded or tunable investor lexicon and dynamic weighting - Source mapping maintenance automation and coverage audits