[mcp-analysis] MCP Structural Analysis - 2026-02-22

[github-actions[bot]]

12 GitHub MCP tools analyzed across 10 toolsets on 2026-02-22. Average usefulness rating: 3.08/5. Two toolsets (context, notifications) are inaccessible in the Actions integration environment (403). Three toolsets (pull_requests, actions, code_security) produce oversized responses requiring a payload file intermediary. Three tools (list_discussions, search_repositories, get_label) deliver compact, high-value responses ideal for agentic use.

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	12
Toolsets Covered	10
Total Tokens (Today)	39,817
Average Usefulness Rating	3.08 / 5
Best Rated Tool	list_discussions — 5/5
Worst Rated Tool	get_me / list_notifications — 1/5 (403 errors)
Most Compact Tool	get_label — 25 tokens
Most Bloated Tool	list_code_scanning_alerts — 25,476 tokens

---

Usefulness Ratings for Agentic Work

Tool	Toolset	Rating	Assessment
list_discussions	discussions	⭐⭐⭐⭐⭐	Compact, clean, proper pagination
search_repositories	search	⭐⭐⭐⭐⭐	Rich metadata in minimal tokens
list_issues	issues	⭐⭐⭐⭐	Complete data; body text inflates size
list_commits	repos	⭐⭐⭐⭐	Clean structure, good nested metadata
get_label	labels	⭐⭐⭐⭐	Flat, minimal, immediately actionable
list_pull_requests	pull_requests	⭐⭐⭐	Oversized — uses payload file system
list_workflows	actions	⭐⭐⭐	Oversized — 205 workflows returned ignoring perPage
list_label	labels	⭐⭐⭐	Oversized for full list; use get_label instead
list_code_scanning_alerts	code_security	⭐⭐	101,905 chars — agent unusable without file read
get_file_contents	repos	⭐⭐	Returns success + SHA only; no file content visible
get_me	context	⭐	403 Not accessible by integration
list_notifications	notifications	⭐	403 Not accessible by integration

---

Schema Analysis

Tool	Type	Depth	Key Fields	Notes
get_me	error	0	—	Inaccessible
get_file_contents	string	1	sha, status	Content not surfaced in response
list_commits	array	3	sha, commit.message, author, committer	Clean nested objects
list_issues	object	3	number, title, state, body, labels, pageInfo	Body can be thousands of chars
list_pull_requests	array	4	id, number, state, title, body, user, head, base	Deep nesting, full body text
list_workflows	object	3	id, name, path, state, total_count	Returns all 205 workflows
list_code_scanning_alerts	array	5	number, rule.id, severity, location, tool	Deeply nested, bloated
list_discussions	object	3	number, title, category.name, user, pageInfo	Ideal compact structure
list_label	object	2	id, color, description, name	Flat list; oversized as full list
get_label	object	1	id, color, description, name	Flat, minimal, perfect
search_repositories	object	3	name, description, language, topics, stats	Rich, compact, excellent
list_notifications	error	0	—	Inaccessible

---

Response Size Analysis

Toolset	Avg Tokens	Tools Tested	Size Category
code_security	25,476	1	🔴 Danger
pull_requests	5,911	1	🔴 Danger
actions	3,885	1	🟠 Warning
labels	1,390	2	🟠 Warning
issues	1,250	1	🟠 Warning
repos	120	2	🟢 Good
search	150	1	🟢 Good
discussions	100	1	🟢 Good
context	13	1	⚫ Error
notifications	13	1	⚫ Error

---

Tool-by-Tool Analysis

Tool	Toolset	Tokens	Schema	Nesting	Rating	Notes
get_me	context	13	error	0	1	403 — not accessible by integration
get_file_contents	repos	14	string	1	2	Returns SHA only; no file content surfaced
list_commits	repos	225	array	3	4	Clean, compact, good metadata
list_issues	issues	1,250	object	3	4	Full body inflates; use field filtering if possible
list_pull_requests	pull_requests	5,911	array	4	3	Payload file intermediary required
list_workflows	actions	3,885	object	3	3	Ignores perPage — returns all 205 workflows
list_code_scanning_alerts	code_security	25,476	array	5	2	101KB response — far too large for direct use
list_discussions	discussions	100	object	3	5	Ideal compact response
list_label	labels	2,755	object	2	3	Use get_label for individual items instead
get_label	labels	25	object	1	4	Clean single-item, flat, minimal
search_repositories	search	150	object	3	5	Rich metadata, topics, stats in one call
list_notifications	notifications	13	error	0	1	403 — not accessible by integration

---

30-Day Trend Summary

Metric	Value
Data Points	12 (first run — no prior history)
Date Range	2026-02-22 (baseline)
Avg Daily Tokens	39,817
Avg Usefulness Rating	3.08 / 5
Inaccessible Tools	2 (get_me, list_notifications)

---

Recommendations

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

• list_discussions — Best balance of data completeness and token efficiency
• search_repositories — Rich multi-field metadata in a single compact call
• list_commits — Good commit metadata with reasonable size
• get_label — Perfect for targeted label lookups

Tools needing improvement:

• list_code_scanning_alerts — No pagination enforcement; returns massive payload even for exploratory calls; needs per_page support and field reduction
• list_workflows — perPage parameter appears ignored; returns all 205 workflows
• get_file_contents — Should surface file content directly, not just a success acknowledgment

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

• get_label (25 tokens, ⭐⭐⭐⭐), list_discussions (100 tokens, ⭐⭐⭐⭐⭐), search_repositories (150 tokens, ⭐⭐⭐⭐⭐)

Context-heavy tools to use cautiously:

• list_code_scanning_alerts (25,476 tokens), list_pull_requests (5,911 tokens), list_workflows (3,885 tokens) — all trigger the payload file intermediary which requires an extra file-read step

---

Visualizations

Tool Response Sizes

Usefulness Ratings by Tool

Toolset: Avg Rating vs Avg Token Size

Token Size vs Usefulness Scatter

References:

• §22279413469

AI generated by GitHub MCP Structural Analysis

• expires on Mar 1, 2026, 3:01 PM UTC

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-03-01T15:01:41.591Z.

Closed by Workflow