Loading ToolBench

Reading the current corpus…

gnews-agent MCP server · D · ToolBench

← Browse

gnews-agent

github.com/ranahaani/gnews-agent

Persistent, semantic, evaluated news intelligence layer for AI agents — built on top of GNews.

2026-07-28+STDIO5 tools

56/100

Definition Quality71

Protocol Readiness47

Supportability38

Spec posture

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

Static source inference · medium confidence · detected: Logging

Deprecated protocol patterns detected

Summary

gnews-agent has well-structured tool definitions with complete schemas and reasonable descriptions. All 5 tools have input schemas with typed parameters and descriptions. However, descriptions are terse (avg ~60 chars, below the 194-char baseline), lack WHEN-to-use guidance, and several tools have underdocumented dependencies. The monitor_topic tool is stateless (v1 limitation) and lacks confirmation patterns for webhook registration. Output schemas are not explicitly documented. Error handling is minimal, no recovery guidance or actionable error messages visible in the code.

Tools (5)

get_briefread onlyauthsource verified70/100

Cited LLM brief on a topic. Requires LLM key.

get_sentimentread onlyauthsource verified70/100

Sentiment scoring for a topic. Requires LLM key.

get_timelineread onlysource verified68/100

Day-by-day article counts (keyless).

monitor_topicwritesource verified67/100

search_newsread onlysource verified80/100

Semantic search over locally-stored articles.

Issues (5)

HIGHtool-description ↗

Descriptions lack WHEN-to-use guidance and prerequisites. 'Semantic search over locally-stored articles' does not explain when to use semantic vs keyword search, or that articles must be pre-ingested. LLMs cannot infer these distinctions.

search_newsget_briefget_sentimentget_timeline

HIGHtool-description ↗

get_brief and get_sentiment require an LLM key but do not document this as a prerequisite or error condition. If the key is missing, the tool will fail with no guidance on how to recover.

get_briefget_sentiment

HIGHtool-description ↗

monitor_topic is stateless (v1 limitation per code comment). The description does not warn that webhook registration is acknowledged but polling is not yet implemented. Agents may expect immediate monitoring and be surprised when no webhooks fire.

monitor_topic

HIGH

Recommendations

Expand tool descriptions to 100 - 200 characters. Add WHEN-to-use guidance: 'Use semantic=true for conceptual queries (e.g., "AI safety"), semantic=false for exact keyword matches (e.g., "OpenAI Q3 earnings").'
Document output schemas in descriptions. E.g., 'Returns list of articles with fields: title (str), url (str), source (str), published_date (ISO 8601), snippet (str).'
Add prerequisite warnings to get_brief and get_sentiment: 'Requires OPENAI_API_KEY or LLM_API_KEY environment variable. If missing, returns error with setup instructions.'
Update monitor_topic description to clarify v1 limitation: 'Registers topic monitor (v1 acknowledges only; polling loop ships in v2). Webhook URL is validated but not yet called.'
Implement error handling with actionable messages. E.g., 'No articles found for query. Try: (1) expanding days parameter, (2) using semantic=false for exact matches, (3) checking country/language filters.'

Spec posture evidence

Inferred effective spec: 2026-07-28+.

Relies on Logging (deprecated) - log to stderr or use OpenTelemetry

Score history

Overall score trend

First recorded score · v2 rubric

56/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	D	56	2026-07-28+	v2