MCP server for GridNews — market news, press releases, and sentiment for AI agents.
GridNews MCP has well-structured tool definitions with comprehensive schemas and clear descriptions. All 9 tools have explicit input schemas with type constraints, enums, and parameter descriptions. Tool names follow verb_noun convention (search_news, get_top_events, list_sources). Descriptions are detailed and contextual (avg ~150 chars), explaining WHEN to use each tool vs alternatives. However, output schemas are not formally documented in the tool definitions, responses are formatted as markdown strings rather than structured JSON objects. Error handling is minimal; no recovery guidance or actionable error messages visible in the code. Security is solid (no secrets in params, read-only operations). Composition is clean: each tool has one responsibility, and tool outputs contain IDs for chaining (e.g., eventId from get_top_events feeds into get_event_detail).
Every outlet that carried one story, grouped by independent voice, with the signal that matched each into the cluster. Use it to check how well-corroborated a story actually is before repeating it, and to cite the original reporting. Event ids are not permanent — clusters are rebuilt as the rules change, so re-resolve from get_top_events rather than storing ids.
Recent news and press releases for a single ticker symbol, plus an aggregate sentiment reading across the recent coverage. Use this when the question is about one company; use search_news when it spans several or needs free-text matching.
Deeper sentiment analysis for one ticker over a timeframe, with a positive/negative/neutral breakdown and the basis it was computed from. Requires pro tier or above. Slower than the other tools because it analyses articles on demand. Check the 'not enough data' note before reporting a reading as meaningful.
The biggest stories right now, as clusters of coverage rather than individual articles. Prefer this over search_news for open-ended questions like 'what is happening in the markets' — it ranks by how many INDEPENDENT outlets corroborated a story, so it surfaces what matters rather than whatever was published most recently. Report the independent-voice count, not the outlet count: outlets that republish each other count once, and a press release is a single issuer however many wires carried it.
Output schemas not formally documented. Tools return markdown strings rather than structured JSON. LLMs cannot reliably parse unstructured text responses or plan downstream tool calls without knowing the response schema.
No error handling guidance. Tools return raw API errors without recovery hints. E.g., 'Requires pro tier or above' appears in descriptions but no tool explains what to do when a tier-gated call fails or how to check entitlements.
Pagination not fully documented in output. search_news and list_press_releases accept page/limit but response structure (total count, has_next, etc.) is not declared in tool definitions.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 66 | 2026-07-28+ | v2 |
Trending topic clusters across recent coverage, as keyword groups with article counts. Useful for answering 'what is the market talking about right now' without a specific query. Works without an API key.
Report the configured API key's tier, feature entitlements, and remaining daily quota. Call this to explain a permission or rate-limit failure from another tool rather than guessing at the cause.
Company press releases straight off the wires (PR Newswire, GlobeNewswire, Business Wire and others), filterable by symbol, provider, company name, and date. Use this when the user specifically wants primary-source company announcements rather than journalism about them.
List the news sources GridNews indexes, with their ids. Call this before using the `sources` filter on search_news so you filter on ids that actually exist.
Search GridNews for market news articles and press releases. The primary tool — use it for any question about what has happened in the markets, at a company, or in a sector. Filter by free text, ticker symbols, source, date range, sentiment, and quality score. Results are newest-first and reranked against the query. Cite the source URL for any claim you repeat.
get_usage description is vague ('Report the configured API key's tier...'). Does not explain when to call it or what fields are returned. Lacks actionable context for LLM selection.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All tools are read-only but this is not formally declared in the protocol, forcing LLMs to infer safety from descriptions.