[mcp-analysis] MCP Structural Analysis - 2026-03-25

[github-actions[bot]]

11 GitHub MCP tools analyzed today across 10 toolsets. Top performers: list_discussions (5/5), get_label (5/5), search_repositories (5/5). Persistent problems: list_code_scanning_alerts remains unusable (~137KB, no pagination), list_workflows ignores per_page parameter, get_me and list_notifications are consistently 403. A notable observation: list_pull_requests response quality improved today — head.repo/base.repo were trimmed to full_name+description only, boosting it to 4/5.

Average usefulness today: 3.09/5 | 30-day average by toolset: search 4.61, discussions 4.59, labels 4.27.

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	11
Total Tokens (Today)	41,452
Avg Usefulness Rating	3.09 / 5
Best Rated Tools	list_discussions, get_label, search_repositories — 5/5
Worst Rated Tools	get_me, list_notifications, list_code_scanning_alerts — 1/5
Analysis Date	2026-03-25
30-Day Data Points	218 records across 22 days

---

Usefulness Ratings for Agentic Work (Today)

Tool	Toolset	Rating	Assessment
list_discussions	discussions	⭐⭐⭐⭐⭐	Compact, cursor pagination, all key fields, perPage respected
get_label	labels	⭐⭐⭐⭐⭐	Minimal flat object (~110 chars), all fields, zero bloat
search_repositories	search	⭐⭐⭐⭐⭐	Rich metadata, topics array, minimal_output=true prevents bloat
list_commits	repos	⭐⭐⭐⭐	Clean compact array, nested author metadata, actionable
list_pull_requests	pull_requests	⭐⭐⭐⭐	Improved today — head/base repo trimmed; body is structured markdown
get_file_contents	repos	⭐⭐⭐	Returns real content now, but README (~8500 chars) heavily inflates context
list_issues	issues	⭐⭐⭐	perPage respected; no cursor visible today; body bloat (~1800 chars)
list_workflows	actions	⭐⭐	per_page=1 ignored (returns 30 of 234); 3 redundant URL fields per item
get_me	context	⭐	Consistently 403 in this integration
list_notifications	notifications	⭐	Consistently 403 in this integration
list_code_scanning_alerts	code_security	⭐	~137KB response, no per_page support, deeply nested — unusable inline

---

Schema Analysis

Tool	Type	Depth	Key Fields	Notes
get_me	error	0	—	403 in this env
get_file_contents	object	2	sha, content.text, resource.uri, mimeType	Full content in resource block; large files inflate context
list_commits	array	3	sha, html_url, commit.message, commit.author, author.login	Clean nested structure
list_issues	array	3	number, title, state, body, labels, user, created_at	No cursor/pageInfo visible in response today
list_pull_requests	array	3	number, title, state, body, head.ref, base.ref, head.repo.full_name, draft	Trimmer head/base than prior days
list_workflows	object	2	id, name, path, state, url, html_url, badge_url, total_count	3 redundant URL fields; pagination ignored
list_code_scanning_alerts	array	5	number, rule.id, rule.severity, state, tool.name, location.path	Includes full rule help text per alert
list_discussions	object	3	number, title, category.name, user.login, created_at, pageInfo, totalCount	Model structure for all tools
get_label	object	1	id, color, description, name	Exemplary minimal response
search_repositories	array	2	full_name, description, language, topics, stargazers_count, forks_count	Topics array is high-value field
list_notifications	error	0	—	403 in this env

---

Response Size Analysis

Toolset	Avg Tokens (30-day)	Today's Tokens	Tools Tested
code_security	~17,000	34,400 (est.)	list_code_scanning_alerts
repos	~460	2,275	get_file_contents, list_commits
pull_requests	~3,270	500	list_pull_requests
actions	~3,350	3,375	list_workflows
issues	~1,970	625	list_issues
labels	~60	28	get_label
discussions	~200	100	list_discussions
search	~390	113	search_repositories
context	~11	18	get_me
notifications	~13	18	list_notifications

---

30-Day Trend Summary

Metric	Value
Total Records	218
Days Tracked	22
Avg Daily Tokens (excl. code_security)	8,706
Most Stable Top Performer	list_discussions — consistently 4–5/5
Most Stable Poor Performer	list_code_scanning_alerts — consistently 1–2/5
Biggest Volatility	list_issues — ranges 2/5 (huge body) to 4/5 (short body)
New Behavior (2026-03-17)	get_file_contents started returning actual file content (was SHA-only)
Today's Notable Change	list_pull_requests head/base repo trimmed → rating improved to 4/5

---

Recommendations

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

• list_discussions, get_label, search_repositories — always compact and actionable
• list_commits — reliable and clean
• list_pull_requests — use with perPage=1; body may still bloat on large PRs

Tools needing pagination discipline:

• list_issues — body content highly variable; consider fetching number/title only then fetching body separately
• list_workflows — per_page ignored; plan for 30-item minimum response; add filtering after fetch

Tools to avoid calling in hot loops:

• list_code_scanning_alerts — skip entirely for context-sensitive agents; use get_code_scanning_alert by number instead
• get_file_contents with large files — README is ~8500 chars; prefer targeted smaller files

Integration limitations (not fixable by agent):

• get_me and list_notifications always return 403 in this integration environment

---

Visualizations

Response Size by Toolset (30-Day Average)

Usefulness Ratings by Toolset (Today)

Daily Token Usage Trend (Excluding code_security)

Token Size vs Usefulness (Today)

Rating Trends by Toolset (30-Day)

References: §23538228262

AI generated by GitHub MCP Structural Analysis · history

• expires on Mar 26, 2026, 11:23 AM UTC

[github-actions[bot]]

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

A newer discussion is available at Discussion #23094.