[mcp-analysis] MCP Structural Analysis - 2026-07-02

[github-actions[bot]]

GitHub MCP Structural Analysis — 2026-07-02

All 10 representative GitHub MCP tools were reachable via the configured HTTP server this run (awmg-github v0.3.32). 10 tools analyzed, avg usefulness 3.60/5, ~10.2K tokens total.

Key findings

• Best-rated (5/5): get_file_contents (full text + SHA, immediately actionable) and list_discussions (only tool with real GraphQL cursor pagination + totalCount).
• Worst-rated (1/5): get_me — still 403 Resource not accessible by integration; the CI GITHUB_TOKEN lacks /user scope. Environmental, not a design regression.
• Cost hotspots: search_repositories (~3969 tok, no perPage → full default page), list_pull_requests (~2878 tok, base/head each re-embed a full repo object), list_issues (~1874 tok, driven by long AI-generated issue bodies).
• Cheapest clean signals: list_code_scanning_alerts returned [] (19 tok — zero open critical/high alerts), get_label (69 tok), search_users (146 tok).
• 30-day trend: avg usefulness rising (first week 3.45 → last week 3.82); the users/actions toolsets are consistently reachable again after earlier permission gaps.

Recommendations

• Prefer (rating 4-5): get_file_contents, list_discussions, list_issues, list_pull_requests, actions_list, search_users.
• Most context-efficient (low tokens, high rating): list_discussions (196 tok/5⭐), actions_list (173 tok/4⭐), search_users (146 tok/4⭐).
• Use with guards (high token cost): always pass perPage to search_repositories; avoid fetching body on list_issues/list_pull_requests when only triaging; expect list_code_scanning_alerts to balloon (~1–2.4K tok/alert from embedded rule.help) on non-empty results — fetch full rule help on-demand instead.
• Avoid on integration tokens: get_me — derive identity from workflow context instead.

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	10
Total Tokens (Today)	10,216
Average Usefulness Rating	3.60/5
Best Rated Tool	get_file_contents & list_discussions: 5/5
Worst Rated Tool	get_me: 1/5 (403 integration)

Usefulness Ratings for Agentic Work

Tool	Toolset	Rating	Assessment
get_file_contents	repos	⭐⭐⭐⭐⭐	Full file text + SHA + uri, immediately actionable for reads/edits
list_discussions	discussions	⭐⭐⭐⭐⭐	Only tool with true cursor pagination + totalCount; compact items
list_issues	issues	⭐⭐⭐⭐	Rich actionable schema; cost driven by long issue bodies, not bloat
list_pull_requests	pull_requests	⭐⭐⭐⭐	Complete PR-ops data; base/head repo objects redundant
actions_list	actions	⭐⭐⭐⭐	Compact workflow metadata + total_count for planning
search_users	users	⭐⭐⭐⭐	Minimal item shape; clean for resolving handles
list_code_scanning_alerts	code_security	⭐⭐⭐	Empty [] here (tiny); very verbose per-alert when non-empty
get_label	labels	⭐⭐⭐	Clean but needs exact label name; list_label better for discovery
search_repositories	search	⭐⭐⭐	Clean minimal_output but unbounded page size (~16KB) without perPage
get_me	context	⭐	403 Resource not accessible by integration — unusable this context

Schema Analysis

Tool	Type	Depth	Key Fields	Notes
get_file_contents	object	3	content, resource.text, mimeType, sha, uri	Resource envelope; SHA enables follow-up writes
list_discussions	object	3	discussions[].{category,number,title,user}, pageInfo, totalCount	GraphQL cursor pagination
list_issues	array	3	number, title, state, labels, user, body, timestamps	Labels as plain names; no cursor
list_pull_requests	array	4	number, state, draft, merged, base, head, assignees, body	base/head embed full repo objects
actions_list	object	3	total_count, workflows[].{id,name,path,state}	total_count but no cursor
search_users	object	3	total_count, items[].{login,id,profile_url,avatar_url}	Compact search envelope
list_code_scanning_alerts	array	1	(empty this run)	Verbose rule.help per item when populated
get_label	object	1	name, color, description, id	Flat single-record lookup
search_repositories	array	3	full_name, language, stars, forks, topics, open_issues_count	No perPage → full page
get_me	error	1	error	403 integration token

Response Size Analysis

Toolset	Avg Tokens	Tools Tested
search	3,969	search_repositories
pull_requests	2,878	list_pull_requests
issues	1,874	list_issues
repos	832	get_file_contents
discussions	196	list_discussions
actions	173	actions_list
users	146	search_users
labels	69	get_label
context	60	get_me
code_security	19	list_code_scanning_alerts

Tool-by-Tool Analysis

Tool	Toolset	Tokens	Schema	Rating	Notes
get_file_contents	repos	832	object/3	5	Full text + SHA-pinned uri; scales with file size, no range window
list_discussions	discussions	196	object/3	5	Cursor pagination + totalCount=7973; no body bloat
list_issues	issues	1,874	array/3	4	perPage=1 but large AI-report body; core fields actionable
list_pull_requests	pull_requests	2,878	array/4	4	Complete; base/head redundant repo objects; body inflates cost
actions_list	actions	173	object/3	4	total_count=331; lean workflow items
search_users	users	146	object/3	4	login/id/profile_url/avatar_url; reachable this run
list_code_scanning_alerts	code_security	19	array/1	3	Empty [] (0 open critical/high); verbose when populated
get_label	labels	69	object/1	3	Clean but needs known name
search_repositories	search	3,969	array/3	3	Unbounded default page; set perPage
get_me	context	60	error/1	1	403 integration token

30-Day Trend Summary

Metric	Value
Data Points	164
Distinct Days	18
Average Daily Tokens	~7,714
Average Rating Trend	Improving (3.45 → 3.82 week-over-week)

Recommendations

• High-value tools (rating 4-5): get_file_contents, list_discussions, list_issues, list_pull_requests, actions_list, search_users
• Tools needing improvement: get_me (403 on integration tokens), search_repositories (unbounded page — enforce perPage), list_code_scanning_alerts (strip/on-demand rule.help to cut per-alert bloat)
• Context-efficient (low tokens, high rating): list_discussions (196/5⭐), actions_list (173/4⭐), search_users (146/4⭐)
• Context-heavy (high tokens): search_repositories (3,969), list_pull_requests (2,878), list_issues (1,874)

Visualizations

Response Size by Toolset

Usefulness Ratings

Daily Token Trend

Size vs Usefulness

References: §28590153883

Warning

Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• awmgmcpg

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "awmgmcpg"

See Network Configuration for more information.

Generated by 🔍 GitHub MCP Structural Analysis · 180.6 AIC · ⌖ 16.3 AIC · ⊞ 9.9K · ◷

• expires on Jul 3, 2026, 4:45 AM UTC-08:00