[mcp-analysis] MCP Structural Analysis - 2026-07-08

[github-actions[bot]]

GitHub MCP Structural Analysis — 2026-07-08

Probed 10 representative GitHub MCP tools (one per toolset) against github/gh-aw with minimal parameters, measuring response size, schema structure, and usefulness for agentic work (1–5).

Key findings today

• Tools analyzed: 10 · Total tokens: 10,074 · Average usefulness: 3.20 / 5
• ⭐ Best rated: actions_list and get_label (5/5) — compact, flat, immediately actionable.
• ⚠️ Worst rated: get_me and list_code_scanning_alerts (1/5) — both returned 401 auth errors under the workflow integration token (environmental limit, not a schema regression).
• 🐘 Heaviest response: list_pull_requests — 5,520 tokens for a single PR (nesting depth 5), dominated by full nested head/base repo objects, requested_reviewers, and _links noise.
• 📉 Efficiency pattern holds: tiny lookups (get_label 58 tok, actions_list 139 tok) rate 5/5, while list endpoints that inline full bodies/nested repos (list_pull_requests, list_issues, search_repositories) carry heavy per-item bloat.

Note on this run's auth context: get_me and list_code_scanning_alerts hit 401 Requires authentication, and list_discussions fell back to a structural estimate (anon GraphQL blocked). These are token/permission limitations of the CI integration identity, consistent with prior runs — with a PAT these tools are normally rated 5, 2–3, and 5 respectively.

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	10
Total Tokens (Today)	10,074
Average Usefulness Rating	3.20 / 5
Best Rated Tool	actions_list, get_label: 5/5
Worst Rated Tool	get_me, list_code_scanning_alerts: 1/5

Usefulness Ratings for Agentic Work

Tool	Toolset	Rating	Assessment
actions_list	actions	⭐⭐⭐⭐⭐	Compact {total_count, workflows[]}; excellent for workflow targeting
get_label	labels	⭐⭐⭐⭐⭐	Flat, minimal, fully actionable lookup
get_file_contents	repos	⭐⭐⭐⭐	Complete file + sha for follow-up edits; scales with file size
list_discussions	discussions	⭐⭐⭐⭐	Normally best-in-class GraphQL cursor pagination (estimate this run)
search_users	users	⭐⭐⭐⭐	Clean {total_count, items[]} handle resolution; no cursor
list_issues	issues	⭐⭐⭐	Crisp triage fields but full body + *_url bloat inlined
search_repositories	search	⭐⭐⭐	All discovery signals but heavy full-repo item; always bound perPage
list_pull_requests	pull_requests	⭐⭐	Extremely complete but 5.5K tokens/PR; nested repo objects redundant
get_me	context	⭐	401 under integration token — unusable here
list_code_scanning_alerts	code_security	⭐	401 (needs GHAS/auth token); no data this run

Schema Analysis

Tool	Type	Depth	Key Fields	Notes
get_me	error	1	message, documentation_url, status	401 error object
get_file_contents	object	2	name, path, sha, size, content, encoding	base64 content drives size
list_issues	array	4	number, title, state, user, labels, created_at	nested user/labels + full body
list_pull_requests	array	5	number, title, state, head, base, labels, draft	full head/base repo objects
actions_list	object	3	total_count, workflows	clean envelope, total_count=331
list_code_scanning_alerts	error	1	message, documentation_url, status	401 error object
list_discussions	array	4	number, title, author, category, createdAt, url	nested author/category (estimate)
get_label	object	1	id, name, color, default, description	flat, minimal
search_users	object	3	total_count, incomplete_results, items	items carry login/id/urls
search_repositories	object	4	total_count, incomplete_results, items	item = full repo object

Response Size Analysis

Toolset	Avg Tokens	Tools Tested
pull_requests	5,520	list_pull_requests
issues	1,637	list_issues
search	1,508	search_repositories
repos	562	get_file_contents
discussions	300	list_discussions
users	296	search_users
actions	139	actions_list
labels	58	get_label
code_security	29	list_code_scanning_alerts (error)
context	25	get_me (error)

Tool-by-Tool Analysis

Tool	Toolset	Tokens	Schema	Rating	Notes
list_pull_requests	pull_requests	5,520	array/d5	⭐⭐	22K chars for ONE PR; nested head/base repo objects + _links noise
list_issues	issues	1,637	array/d4	⭐⭐⭐	Full body + many *_url fields inflate a single issue
search_repositories	search	1,508	object/d4	⭐⭐⭐	Single item = full repo object; total_count=778
get_file_contents	repos	562	object/d2	⭐⭐⭐⭐	LICENSE via base64 content + sha; actionable once decoded
list_discussions	discussions	300	array/d4	⭐⭐⭐⭐	Estimate; nested author/category, normally cursor-paginated
search_users	users	296	object/d3	⭐⭐⭐⭐	total_count=2; compact items[]; param name is query
actions_list	actions	139	object/d3	⭐⭐⭐⭐⭐	total_count=331; needs method=list_workflows
get_label	labels	58	object/d1	⭐⭐⭐⭐⭐	Flat bug-label object; best token efficiency
list_code_scanning_alerts	code_security	29	error/d1	⭐	401 requires-auth; verbose (rule.help) when authorized
get_me	context	25	error/d1	⭐	401 requires-auth; read actor from github-context instead

30-Day Trend Summary

Metric	Value
Data Points	166
Distinct Days	18 (2026-06-08 → 2026-07-08)
Average Daily Total Tokens	~7,658
Average Rating Trend	Stable (3.22 → 3.20 over window)

Recommendations

• High-value tools (rating 4–5): get_label, actions_list, get_file_contents, list_discussions, search_users — prefer these for cheap, actionable reads.
• Tools needing improvement: list_pull_requests (strip nested head/base repo objects and _links); list_issues (excerpt bodies instead of dumping them).
• Context-efficient tools (low tokens, high rating): get_label (58 tok, 5/5), actions_list (139 tok, 5/5) — ideal defaults for discovery/lookup.
• Context-heavy tools (high tokens): list_pull_requests (5.5K/PR), list_issues (1.6K/issue), search_repositories (1.5K) — always set perPage and avoid list endpoints when a targeted get_* will do.
• Environmental: get_me and list_code_scanning_alerts are unusable under the CI integration token (401). Derive identity from github-context; run code-scanning reads with a GHAS-scoped token.

Visualizations

Response Size by Toolset

Usefulness Ratings

Daily Token Trend

Size vs Usefulness

References: §28942201862

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 · 186.4 AIC · ⌖ 16.3 AIC · ⊞ 9.6K · ◷

• expires on Jul 9, 2026, 4:36 AM UTC-08:00

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-07-09T12:36:28.410Z.

Closed by Workflow