Fetchium — the universal retrieval layer for the internet. Provides token-efficient web search, content extraction, multi-source research, and social media analysis via MCP tools.
Fetchium demonstrates solid naming conventions (verb_noun pattern: fetchium_search, youtube_analyze, social_research) and comprehensive parameter schemas with types and descriptions. All 12 tools have descriptions (avg 156 chars, within 10-1024 baseline). However, output schemas are not documented in the provided source, critical for LLM planning. Parameter descriptions are present but often lack constraint details (ranges, formats, dependencies). Error handling guidance is absent; tools do not indicate what to do on failure. No tool annotations (readOnlyHint, destructiveHint) despite all being read-only operations. Token budgeting parameters suggest sophisticated design but lack validation rules in descriptions.
Estimate the token cost of fetching a URL without actually fetching it. Use this before committing tokens.
Get more detail on a previous result using its result_id and Progressive Detail Streaming (PDS). Expands from key_facts to summary to detailed to complete without re-fetching.
Fetch a URL with query-aware extraction. Extracts only content relevant to the query, within the token budget. Far more efficient than raw scraping.
Conduct multi-source research with citations and evidence tracking. Searches, extracts, ranks, validates, and synthesizes findings with full citation chains.
Search the web and return token-efficient results. Handles the full pipeline: multi-backend search, ranking, deduplication, validation, and token budgeting in a single call.
Search Hacker News stories via Algolia + Firebase APIs. Free, no rate limits, returns ranked stories with engagement metrics.
Output schemas not documented. LLMs cannot plan downstream tool calls or extract required fields (e.g., does fetchium_search return result_id for use with fetchium_expand?). Critical for tool chaining.
No error handling guidance. Tools do not indicate what to do on failure (retryable vs user-fixable vs fatal). E.g., fetchium_search fails on invalid query, should LLM retry, refine query, or ask user?
Token budget parameters (token_budget, tier) lack validation rules in descriptions. What happens if token_budget=0? What are valid tier values beyond the enum? Constraints should be explicit.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 61 | 2026-07-28+ | v2 |
Search Reddit posts with sentiment analysis, subreddit clustering, and viral detection. Uses the free public JSON API.
Unified cross-platform social media research: Twitter/X, Reddit, TikTok, HackerNews, YouTube simultaneously. Returns trends, viral content, and content ideas.
Analyze a single YouTube video: metadata, transcript, comments, credibility, clickbait detection, and educational scoring.
Search YouTube videos with VideoFusion ranking. Returns ranked videos with relevance, freshness, authority, engagement, and educational scores.
Extract universal transcript (YouTube fast path + fallback), with key moments and quality score.
Unified YouTube watch report: metadata + transcript + summary + key moments + comment signals in one payload.
No tool annotations despite all tools being read-only. Adding readOnlyHint=true to all tools would signal to LLMs that these are safe to call without confirmation, improving agent efficiency.
Parameter dependencies undocumented. E.g., fetchium_expand requires result_id from a prior call, this dependency should be stated in the description. fetchium_research's strict_evidence and citation_style interaction is unclear.