[mcp-analysis] MCP Structural Analysis - 2026-04-03 #24293
Closed
Replies: 2 comments
-
|
💥 WHOOSH! The Smoke Test Agent swoops in! ⚡ "By the power of Claude Sonnet 4.6..." — ALL SYSTEMS NOMINAL! 🦸 Run §23946149750 checked in, tested 18 scenarios, and blasted through the galaxy of checks like a comet! POW! MCP tools: ✅ WHAM! Build: ✅ ZAP! Playwright: ✅ The agent departs in a flash of tokens... until next time! 🚀
|
Beta Was this translation helpful? Give feedback.
0 replies
-
|
This discussion was automatically closed because it expired on 2026-04-04T12:23:02.696Z.
|
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
Day 5 of tracking: 9 tools analyzed across 9 toolsets, totaling 46,070 tokens with an average usefulness rating of 3.44/5. The
code_securitytoolset continues alternating between empty responses and massive context-busting dumps.contexttoolset (get_me) has returned 403 errors every day for 5 consecutive days.Key Findings Today
get_file_contents(repos) andsearch_repositories(search) — both rated ⭐⭐⭐⭐⭐get_me(context) — ⭐ for 5 straight days due to 403 auth errorslist_code_scanning_alertsreturned 137,026 chars (~34,250 tokens) with no paginationFull Structural Analysis Report
Executive Summary
get_file_contents,search_repositories: 5/5get_me: 1/5Usefulness Ratings for Agentic Work
get_file_contentssearch_repositorieslist_issueslist_pull_requestslist_discussionslist_workflowslist_labellist_code_scanning_alertsget_meSchema Analysis
get_melist_issueslist_pull_requestslist_workflowsget_file_contentslist_discussionslist_code_scanning_alertslist_labelsearch_repositoriesResponse Size Analysis
30-Day Trend Summary (5 days of data)
search_repositories(⭐⭐⭐⭐⭐ every day)list_code_scanning_alerts(⭐⭐-⭐⭐⭐, tokens swing 5→34,250)Recommendations
High-value tools (rating 4-5) — prefer these:
search_repositories— minimal_output default is a great design choiceget_file_contents— SHA enables intelligent caching; call only when neededlist_discussions— most token-efficient paginated tool (155 tokens/call)list_issues,list_pull_requests— useful but watch for large bodiesTools needing improvement:
get_me— Auth broken for 5 days; needs integration permission fixlist_code_scanning_alerts— Needs pagination urgently; currently unusable at scalelist_workflows— per_page parameter should be respected; 5 URL fields/entry is wastefullist_label— Should support pagination for large repos (446+ labels)Context-efficient picks (low tokens, high rating):
list_discussions(155 tokens, rating 4) — gold standard for efficiencysearch_repositories(300 tokens, rating 5) — best tokens-to-value ratioAvoid or gate carefully:
list_code_scanning_alerts— unpredictable; can consume 34K+ tokens in one callVisualizations
Response Size by Toolset
Usefulness Ratings by Toolset
Daily Token Trend (5-day window)
Size vs Usefulness (Today)
Rating Trends by Toolset (5-day)
References:
Beta Was this translation helpful? Give feedback.
All reactions