[mcp-analysis] MCP Structural Analysis - 2026-04-02

[github-actions[bot]]

Today's analysis tested 9 GitHub MCP tools across 8 toolsets. Day 4 of tracking. Total tokens consumed today: 11,225 (avg 3.56/5 usefulness). Top performers remain get_file_contents, search_repositories (⭐⭐⭐⭐⭐). get_me continues returning 403. list_code_scanning_alerts was empty today (alternates between empty and massive dumps across days — the core pagination problem persists). The 4-day trend shows token usage stabilizing after a spike on day 3.

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	9
Total Tokens (Today)	11,225
Average Usefulness Rating	3.56 / 5
Best Rated Tools	get_file_contents, search_repositories — 5/5
Worst Rated Tool	get_me — 1/5 (403 error, 4 days consistent)
Days of Trend Data	4 (2026-03-30 → 2026-04-02)

---

Usefulness Ratings for Agentic Work

Tool	Toolset	Rating	Assessment
get_file_contents	repos	⭐⭐⭐⭐⭐	Excellent — direct content + SHA; immediately actionable
search_repositories	search	⭐⭐⭐⭐⭐	Excellent — minimal_output default; perfect for discovery
list_discussions	discussions	⭐⭐⭐⭐	Good — very compact (155 tokens), cursor pagination works
list_issues	issues	⭐⭐⭐⭐	Good — body can balloon with diffs; otherwise solid
list_pull_requests	pull_requests	⭐⭐⭐⭐	Good — nested base/head adds depth but draft/merged flags are high-value
list_workflows	actions	⭐⭐⭐	Adequate — per_page ignored; 5 URL fields per entry is bloat
list_label	labels	⭐⭐⭐	Adequate — returns all 445 labels with no pagination; flat but heavy
list_code_scanning_alerts	code_security	⭐⭐⭐	Adequate when empty; massive (34K+ tokens) when alerts exist with no pagination control
get_me	context	⭐	Poor — 403 error in this integration; no value

---

Schema Analysis

Tool	Type	Depth	Key Fields	Notes
get_me	error	1	—	403 in this integration
list_issues	array	3	number, title, state, body, labels, user	body field has no size limit
list_pull_requests	array	4	number, title, state, body, base, head, user, draft	nested repo objects in base/head
list_workflows	object	3	id, name, path, state, created_at, html_url, badge_url	5 URL fields per entry; per_page ignored
get_file_contents	object	2	sha, mimeType, text	clean; SHA enables caching
list_discussions	object	3	number, title, category, user, created_at, pageInfo	pageInfo cursor pagination present
list_code_scanning_alerts	array	4	number, rule, severity, state, location, tool	no pagination; size varies 5–34,584 tokens
list_label	object	2	id, name, color, description	flat; all 445 labels returned; no pagination
search_repositories	array	2	name, full_name, description, language, topics, stars	minimal_output=true by default

---

Response Size Analysis (Today)

Toolset	Tokens	Notes
repos	3,500	README; stable size
actions	3,500	30 workflows; per_page ignored
labels	2,200	All 445 labels
issues	900	Variable (diff in body today)
pull_requests	600	Nested but manageable
search	350	Minimal output — efficient
discussions	155	Best context efficiency
code_security	5	Empty today
context	15	Error response
Total	11,225

---

Tool-by-Tool Analysis

Tool	Toolset	Tokens	Schema	Rating	Notes
get_me	context	15	error	1	403 consistently
list_issues	issues	900	array/3	4	Body had diff today
list_pull_requests	pull_requests	600	array/4	4	Good nested data
list_workflows	actions	3,500	object/3	3	per_page ignored
get_file_contents	repos	3,500	object/2	5	SHA unchanged from Apr 1
list_discussions	discussions	155	object/3	4	totalCount=4427 (+49/day)
list_code_scanning_alerts	code_security	5	array/1	3	Empty; pagination gap persists
list_label	labels	2,200	object/2	3	445 labels; no pagination
search_repositories	search	350	array/2	5	github-mcp-server: 28,481 ⭐

---

4-Day Trend Summary

Date	Total Tokens	Avg Usefulness	Notable
2026-03-30	42,469	3.56	First run; code_security bloat (34K)
2026-03-31	10,015	3.67	Code security empty
2026-04-01	47,039	3.56	Code security bloat again (34K)
2026-04-02	11,225	3.56	Code security empty; stable day

Observations:

• Token usage alternates dramatically based on list_code_scanning_alerts (empty vs full dump)
• Usefulness ratings are stable at 3.56/5 average across all days
• Discussion count growing consistently (~49 new per day)
• README SHA has not changed in 2 days

---

Recommendations

High-value tools (4-5 ⭐) — use freely:

• get_file_contents, search_repositories — gold standard for context efficiency
• list_discussions, list_issues, list_pull_requests — good with awareness of body size

Tools needing improvement:

• list_code_scanning_alerts — critical gap: needs pagination (per_page param); alternates between 0 and 34K+ tokens
• list_workflows — per_page parameter is silently ignored; 5 URL fields per entry is unnecessary
• list_label — no pagination; returns all labels regardless of repo size

Context efficiency tips:

• Prefer search_repositories over get_file_contents for repo discovery
• For issues/PRs, filter by state/label to reduce body bloat
• Avoid list_code_scanning_alerts without defensive context guards

---

Visualizations

Response Size by Tool

Usefulness Ratings

Daily Token Trend

Size vs Usefulness

4-Day Usefulness Heatmap

References:

• §23897788655

AI generated by GitHub MCP Structural Analysis · history

• expires on Apr 3, 2026, 11:25 AM UTC

[github-actions[bot]]

This discussion has been marked as outdated by GitHub MCP Structural Analysis.

A newer discussion is available at Discussion #24293.