[mcp-analysis] MCP Structural Analysis - 2026-02-23

[github-actions[bot]]

Analysis of 11 GitHub MCP tools across 10 toolsets tested on 2026-02-23. Average usefulness rating: 3.09/5. Best: list_discussions (5/5), Worst: get_me + list_notifications (both 403 errors, 1/5). Two days of trend data now available showing consistent patterns: discussions and search toolsets are highly efficient, while code_security and pull_requests are context-heavy.

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	11
Total Tokens Today	23,811
Average Usefulness Rating	3.09 / 5
Best Rated Tool	list_discussions: 5/5
Worst Rated Tool	get_me, list_notifications: 1/5 (403 errors)
Context-Heaviest Tool	list_code_scanning_alerts: 13,515 tokens
Most Efficient	get_label: 25 tokens, 4/5 rating

---

Usefulness Ratings for Agentic Work

Tool	Toolset	Rating	Assessment
list_discussions	discussions	⭐⭐⭐⭐⭐	Excellent — compact pagination, all key fields, immediately actionable
search_repositories	search	⭐⭐⭐⭐⭐	Excellent — rich metadata (topics, stars, language) in one call
list_issues	issues	⭐⭐⭐⭐	Good — full pagination support; body size varies and can inflate tokens
list_commits	repos	⭐⭐⭐⭐	Good — clean nested structure, compact, good commit metadata
get_label	labels	⭐⭐⭐⭐	Good — minimal flat object, all fields present, low overhead
list_pull_requests	pull_requests	⭐⭐⭐	Adequate — oversized (21,085 chars), requires payload file read step
list_workflows	actions	⭐⭐⭐	Adequate — ignores per_page param, returns all 205 workflows
get_file_contents	repos	⭐⭐	Limited — returns only SHA + success message; no actual content
list_code_scanning_alerts	code_security	⭐⭐	Limited — 54,058 chars, 5-level nesting, no per_page support
get_me	context	⭐	Unavailable — 403 error in this integration environment
list_notifications	notifications	⭐	Unavailable — 403 error in this integration environment

---

Schema Analysis

Tool	Type	Depth	Key Fields	Notes
get_me	error	0	—	403 Not accessible by integration
get_file_contents	string	1	sha, status	No file content returned
list_commits	array	3	sha, html_url, commit.message, commit.author	Clean and compact
list_issues	object	3	number, title, state, body, labels, user, pageInfo	Body can be very large
list_pull_requests	array	4	id, number, state, title, body, user, labels, head, base	Payload file offload
list_workflows	object	3	id, name, path, state, created_at, updated_at	Payload file offload
list_code_scanning_alerts	array	5	number, rule.id, rule.severity, state, tool.name, location.path	Deeply nested, huge
list_discussions	object	3	number, title, category.name, user.login, created_at, pageInfo	Best-in-class structure
get_label	object	1	id, color, description, name	Ideal flat schema
search_repositories	object	3	total_count, items.name, topics, stargazers_count	Rich, well-balanced
list_notifications	error	0	—	403 Not accessible by integration

---

Response Size Analysis

Toolset	Tokens (Today)	Δ vs Yesterday	Tool Tested
code_security	13,515	-47% (101,905→54,058 chars)	list_code_scanning_alerts
pull_requests	5,271	-11%	list_pull_requests
actions	3,885	±0%	list_workflows
issues	600	-52%	list_issues (shorter body today)
repos (commits)	225	±0%	list_commits
search	150	±0%	search_repositories
discussions	100	±0%	list_discussions
labels	25	±0%	get_label
repos (files)	14	±0%	get_file_contents
context	13	±0%	get_me (error)
notifications	13	±0%	list_notifications (error)

---

Tool-by-Tool Analysis

Tool	Toolset	Tokens	Schema	Depth	Rating
list_code_scanning_alerts	code_security	13,515	array	5	2/5
list_pull_requests	pull_requests	5,271	array	4	3/5
list_workflows	actions	3,885	object	3	3/5
list_issues	issues	600	object	3	4/5
list_commits	repos	225	array	3	4/5
search_repositories	search	150	object	3	5/5
list_discussions	discussions	100	object	3	5/5
get_label	labels	25	object	1	4/5
get_file_contents	repos	14	string	1	2/5
get_me	context	13	error	0	1/5
list_notifications	notifications	13	error	0	1/5

---

30-Day Trend Summary

Metric	Value
Data Points	23
Days with Data	2 (2026-02-22, 2026-02-23)
Total Tokens Day 1	33,812
Total Tokens Day 2	23,811
Change	-30% (driven by fewer code scanning alerts)
Average Rating (Day 1)	3.09/5
Average Rating (Day 2)	3.09/5
Rating Trend	Stable

---

Recommendations

High-value tools (rating 4–5, low tokens): list_discussions, search_repositories, get_label, list_commits, list_issues

Tools needing improvement:

• list_code_scanning_alerts — add per_page parameter support; current response can exceed 50K chars
• list_workflows — per_page parameter is ignored; always returns full list (205 workflows)
• list_pull_requests — extremely large body content should be truncated or excluded by default
• get_file_contents — should return actual file content, not just SHA confirmation

Context-efficient sweet spot (≤500 tokens + rating ≥4): get_label (25 tokens), list_discussions (100 tokens), search_repositories (150 tokens), list_commits (225 tokens)

Integration environment gaps (403 errors): get_me, list_notifications — these toolsets are unavailable under the current GitHub Actions integration token scope

---

Visualizations

Response Size by Toolset

Usefulness Ratings per Tool

Daily Token Usage Trend

Token Size vs Usefulness (Efficiency Map)

References: §22303777102

AI generated by GitHub MCP Structural Analysis

• expires on Feb 24, 2026, 11:21 AM UTC

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-02-24T11:21:53.008Z.

Closed by Workflow