Running an open catalog scan across popular MCP servers. tavily-mcp came back F-31 — main issues: search tool missing result_count param description, extract tool has no output schema context.
Full report:
Tavily MCP — Twig Audit
| Tool |
Description |
Issues |
| tavily_search |
"Search the web for current information..." |
✅ Decent. Missing: when to prefer over tavily_research; no mention of latency tradeoff |
| tavily_extract |
"Extract content from URLs." |
❌ 19 chars. Catastrophically short. No output context, no error states |
| tavily_crawl |
"Crawl a website starting from a URL." |
❌ No output context, no use-case differentiation from extract |
| tavily_map |
"Map a website" |
❌ 3 words. Zero information. Agents will never call this |
| tavily_research |
"Perform comprehensive research..." |
⚠️ OK but no differentiation from tavily_search |
Score: D-38 / 100
Tavily MCP — Twig Score: D (38/100)
🔍 Analyzing: tavily-mcp (v0.2.18, 5 tools)
tavily_search C+ Description adequate but missing: when to prefer
over tavily_research, no latency/cost tradeoff guidance.
tavily_extract F "Extract content from URLs." — 29 chars. No output
format context, no error states, no guidance on when
basic vs advanced extraction is appropriate.
tavily_crawl F "Crawl a website starting from a URL." — 37 chars.
No output schema context. Agents can't distinguish
from tavily_extract based on description alone.
tavily_map F "Map a website" — 14 chars. Agents will skip this
entirely. No output type, no use case, nothing.
tavily_research C Description functional but no differentiation from
tavily_search — agents will guess between them.
SCORE: 38/100 (D)
Issues: 3 tools with near-zero descriptions, 0 tools with routing guidance between overlapping capabilities.
Just sharing the data. Happy to discuss if useful.
Running an open catalog scan across popular MCP servers. tavily-mcp came back F-31 — main issues: search tool missing result_count param description, extract tool has no output schema context.
Full report:
Tavily MCP — Twig Audit
Score: D-38 / 100
Tavily MCP — Twig Score: D (38/100)
🔍 Analyzing: tavily-mcp (v0.2.18, 5 tools)
tavily_search C+ Description adequate but missing: when to prefer
over tavily_research, no latency/cost tradeoff guidance.
tavily_extract F "Extract content from URLs." — 29 chars. No output
format context, no error states, no guidance on when
basic vs advanced extraction is appropriate.
tavily_crawl F "Crawl a website starting from a URL." — 37 chars.
No output schema context. Agents can't distinguish
from tavily_extract based on description alone.
tavily_map F "Map a website" — 14 chars. Agents will skip this
entirely. No output type, no use case, nothing.
tavily_research C Description functional but no differentiation from
tavily_search — agents will guess between them.
SCORE: 38/100 (D)
Issues: 3 tools with near-zero descriptions, 0 tools with routing guidance between overlapping capabilities.
Just sharing the data. Happy to discuss if useful.