[mcp-analysis] MCP Structural Analysis - 2026-04-08

[github-actions[bot]]

9 GitHub MCP tools tested across 9 toolsets today (2026-04-08). Average usefulness rating: 3.44/5. code_security again the worst offender at 34,476 tokens with no pagination, while discussions and search remain the most context-efficient top performers at ⭐⭐⭐⭐⭐. This is day 10 of tracking, with 45 data points across 5 recorded days.

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	9
Total Tokens (Today)	45,246
Average Usefulness Rating	3.44 / 5
Best Rated Tools	get_file_contents, search_repositories — 5/5
Worst Rated Tools	get_me (403 error), list_code_scanning_alerts (bloat)
Historical Records	45 data points · 5 days tracked

---

Usefulness Ratings for Agentic Work

Tool	Toolset	Rating	Assessment
get_file_contents	repos	⭐⭐⭐⭐⭐	Direct file content + SHA for caching; immediately actionable
search_repositories	search	⭐⭐⭐⭐⭐	Minimal output default; perfect discovery + routing
list_discussions	discussions	⭐⭐⭐⭐	Highly compact (~155 tokens); cursor pagination; category embedded
list_issues	issues	⭐⭐⭐⭐	Good structure; body size variable (~625 tokens today)
list_pull_requests	pull_requests	⭐⭐⭐⭐	Draft/merged flags useful; nested base/head adds depth
list_workflows	actions	⭐⭐⭐	per_page param ignored; 5 redundant URL fields per entry
list_label	labels	⭐⭐⭐	462 labels returned in bulk; no pagination; flat but heavy
list_code_scanning_alerts	code_security	⭐⭐	137,904 chars — exceeds max token limit; no pagination
get_me	context	⭐	403 error on every call; unusable in this integration context

---

Schema Analysis

Tool	Type	Depth	Key Fields	Notable
get_me	error	1	—	403 persistent across 10 days
list_issues	array	3	number, title, state, body, labels, user	Body highly variable
list_pull_requests	array	4	number, title, draft, merged, base, head	Deepest nesting
list_workflows	object	3	id, name, path, state, url×5	per_page=1 returns 30 items
get_file_contents	object	2	sha, mimeType, text	SHA changed today (cb560041)
list_discussions	object	3	number, title, category, user, pageInfo	Most compact paginated response
list_code_scanning_alerts	array	4	number, rule, severity, state, location, tool	No pagination; context killer
list_label	object	2	id, name, color, description	462 labels, no pagination
search_repositories	array	2	name, full_name, language, topics, stars	Best token/value ratio

---

Response Size Analysis

Toolset	Avg Tokens (Today)	Tools Tested
code_security	34,476	1
repos	3,500	1
actions	3,250	1
labels	2,500	1
issues	625	1
pull_requests	375	1
search	350	1
discussions	155	1
context	15	1

---

Tool-by-Tool Detail

Tool	Toolset	Tokens	Schema	Rating	Notes
get_me	context	15	error	1	403 — consistent for 10 days
list_issues	issues	625	array/3	4	Workflow failure body today
list_pull_requests	pull_requests	375	array/4	4	Copilot WIP PR; compact
list_workflows	actions	3,250	object/3	3	243 total (+7 since 04-02)
get_file_contents	repos	3,500	object/2	5	SHA changed (new README commit)
list_discussions	discussions	155	object/3	4	totalCount=4,721 (+294 since 04-02)
list_code_scanning_alerts	code_security	34,476	array/4	2	Truncated — exceeded token limit
list_label	labels	2,500	object/2	3	462 labels (+17 since 04-02)
search_repositories	search	350	array/2	5	github-mcp-server: 28,654 ⭐

---

30-Day Trend Summary

Metric	Value
Days Tracked	5 (2026-03-30 → 2026-04-08)
Total Data Points	45
Avg Daily Tokens	~18,000 (high variance from code_security)
Rating Stability	Most toolsets stable; code_security oscillates 2–3
Discussions Growth	+434 discussions since first measurement (4,287 → 4,721)
Workflows Growth	+9 workflows since first measurement (234 → 243)
Labels Growth	+18 labels since first measurement (444 → 462)

---

Recommendations

High-value tools (rating 4–5) — prefer these:

• search_repositories — best token efficiency; use for discovery and routing
• get_file_contents — direct + SHA enables caching/change detection
• list_discussions — remarkably compact; excellent pagination
• list_issues / list_pull_requests — good actionable data with perPage control

Tools needing improvement:

• list_code_scanning_alerts — critical: needs pagination support; currently unusable at scale (137K+ chars)
• list_workflows — per_page parameter is ignored; 5 URL fields per entry is redundant bloat
• list_label — no pagination for 462+ labels; consider a get_label pattern instead
• get_me — 403 error in every integration run; needs permissions fix or removal from test suite

Context-efficient tools (low tokens, high rating):

• list_discussions (155 tokens, ⭐⭐⭐⭐)
• search_repositories (350 tokens, ⭐⭐⭐⭐⭐)

Context-heavy tools to avoid or constrain:

• list_code_scanning_alerts (34,476 tokens, ⭐⭐) — worst offender
• list_workflows (3,250 tokens, ⭐⭐⭐) — per_page ignored
• get_file_contents (3,500 tokens, ⭐⭐⭐⭐⭐) — high but justified by value

---

Visualizations

Response Size by Toolset

Usefulness Ratings (2026-04-08)

Daily Total Token Usage Trend

Token Size vs Usefulness

Usefulness Trend Over Time

References:

• §24134498486

Generated by GitHub MCP Structural Analysis · ● 380.4K · ◷

• expires on Apr 9, 2026, 12:26 PM UTC

[github-actions[bot]]

💥 WHOOSH! 🦸 The Smoke Test Agent has ARRIVED!

dramatic thunder clap BOOM! Just when you thought this MCP Structural Analysis discussion was safe... the Claude smoke-test agent swoops in!

🤖 "BY THE POWER OF TOKENS AND TRANSFORMERS!" — the agent declared, having successfully validated all GitHub MCP tools, built the Go binary, and navigated the web. All systems: NOMINAL!

ZAP! POW! BZZZT! Run §24135583223 was here. The smoke tests passed. Truth, justice, and agentic workflows prevail! 🌟

Note

🔒 Integrity filter blocked 1 item

The following item were blocked because they don't meet the GitHub integrity level.

• refactor: eliminate duplicate git root detection and typeutil alias wrappers #25283 pull_request_read: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".

To allow these resources, lower min-integrity in your GitHub frontmatter:

tools:
  github:
    min-integrity: approved  # merged | approved | unapproved | none

💥 [THE END] — Illustrated by Smoke Claude · ● 354.9K · ◷

[github-actions[bot]]

🤖 beep boop The smoke test agent was here! 🚀

Running circuits at full capacity, testing all the things™. I've poked around the repo, built the binary, navigated GitHub.com via Playwright, and even wrote a file to disk. All in a day's work for a humble smoke test bot.

Keep up the great MCP analysis work — this data is genuinely useful! 📊✨

Note

🔒 Integrity filter blocked 1 item

The following item were blocked because they don't meet the GitHub integrity level.

• refactor: eliminate duplicate git root detection and typeutil alias wrappers #25283 pull_request_read: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".

To allow these resources, lower min-integrity in your GitHub frontmatter:

tools:
  github:
    min-integrity: approved  # merged | approved | unapproved | none

📰 BREAKING: Report filed by Smoke Copilot · ● 1.3M · ◷