[mcp-analysis] MCP Structural Analysis - 2026-09-23

[github-actions[bot]]

MCP Structural Analysis — 2026-09-23

Tested 9 representative GitHub MCP tools plus the injected workflow-context block across all available toolsets. Ratings held stable (avg 3.6/5, in line with the 17-day trend range of 3.2–3.7). The standout finding remains list_code_scanning_alerts: even with a bounded perPage=5 call it cost 5,900 tokens — by far the most expensive single call today — driven by verbatim-repeated rule descriptions rather than alert diversity. Best value tools were the zero-cost github-context block and the clean, wrapper-based list_workflows.

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	10 (9 MCP calls + 1 injected context block)
Total Tokens (Today)	9,868
Average Usefulness Rating	3.6/5
Best Rated Tool	github_context / list_workflows: 5/5
Worst Rated Tool	list_issues: 2/5

Usefulness Ratings for Agentic Work

Tool	Toolset	Rating	Assessment
github_context	workflow_context	⭐⭐⭐⭐⭐	Zero-cost, immediately actionable identity data
list_workflows	actions	⭐⭐⭐⭐⭐	Clean wrapper output, explicit pagination, no bloat
get_file_contents	repos	⭐⭐⭐⭐	Complete file content; icon overhead dominates small files
list_pull_requests	pull_requests	⭐⭐⭐⭐	Flat, minimal, directly actionable
search_code	search	⭐⭐⭐⭐	Inline match fragments avoid follow-up reads
list_code_scanning_alerts	code_security	⭐⭐⭐	Complete but boilerplate-heavy; perPage keeps it bounded
get_label	labels	⭐⭐⭐	Tiny payload, dominated by fixed icon block
list_discussions	discussions	⭐⭐⭐	Good pagination metadata, icon overhead persists
search_users	users	⭐⭐⭐	Minimal identity fields, needs follow-up for detail
list_issues	issues	⭐⭐	Result redacted by integrity policy, blocks assessment

Schema Analysis

Tool	Type	Depth	Key Fields	Notes
github_context	markdown	1	actor, repository, workspace, workflow-run-id	Prompt-injected, no MCP overhead
get_file_contents	object	3	content.resource.text, _meta.serverInfo	Fixed ~2235-char icon block on every call
list_issues	array	2	number, title, state, labels	Empty + 1 integrity-redacted item
list_pull_requests	array	1	number, title, state, html_url	No icon/_meta overhead
list_workflows	object	2	total_count, workflows[].id/name/path/state	mcpscripts wrapper, clean JSON
list_code_scanning_alerts	array	3	rule.id, rule.full_description, rule.help, most_recent_instance.location	Verbatim-repeated boilerplate per instance
list_discussions	object	3	discussions[].category/user, pageInfo, totalCount	Good relational data, icon overhead
get_label	object	1	name, color, description, id	Trivial payload, icon-dominated
search_users	object	2	items[].login/id/profile_url, total_count	Identity-only, no bio/name
search_code	object	2	items[].path, text_matches[].fragment, total_count	Includes matching snippet inline

Response Size Analysis

Toolset	Avg Tokens (17-day)	Tools Tested
code_security	6,830.6	1
repos	2,482.3	1
labels	832.9	1
discussions	695.2	1
users	617.8	1
search	616.3	1
actions	171.2	1
issues	93.8	1
workflow_context	91.6	1
pull_requests	90.1	1

Tool-by-Tool Analysis (Today)

Tool	Toolset	Tokens	Schema	Rating	Notes
github_context	workflow_context	90	markdown	5	Zero-cost identity source
list_workflows	actions	172	object	5	mcpscripts wrapper, total_count=376
get_file_contents	repos	907	object	4	LICENSE file, icon overhead ~60% of payload
list_pull_requests	pull_requests	46	array	4	Single Copilot SDK profile PR
search_code	search	743	object	4	No rate-limiting today
list_code_scanning_alerts	code_security	5,900	array	3	perPage=5 kept it bounded vs. full 30-alert pull
get_label	labels	605	object	3	'bug' label, icon-dominated
list_discussions	discussions	720	object	3	totalCount 10,855, up from 10,821
search_users	users	620	object	3	Identity-only fields
list_issues	issues	65	array	2	Integrity-redacted item #62932

30-Day Trend Summary

Metric	Value
Data Points	170 (17 days × 10 tools)
Average Daily Tokens	12,521.8
Average Rating Trend	Stable (range 3.2–3.7, no drift)

Recommendations

• High-value tools (rating 4–5): github_context, list_workflows (mcpscripts wrapper), get_file_contents, list_pull_requests, search_code.
• Tools needing improvement: list_code_scanning_alerts (deduplicate rule-level full_description/help text from per-instance data — this alone would cut the toolset's average from ~6,800 tokens dramatically); list_issues (integrity-policy redaction blocks agent assessment on every measured day, 17/17).
• Context-efficient tools (low tokens, high rating): github_context (90 tokens, zero MCP cost), list_workflows via mcpscripts wrapper (172 tokens, full pagination metadata).
• Context-heavy tools: list_code_scanning_alerts remains the single largest per-call cost by a wide margin, even when bounded with perPage=5; get_file_contents varies wildly by target file (207x range observed: 46 tokens for a tiny PR list vs 36,600+ for unbounded alert pulls).

Visualizations

Response Size by Toolset

Usefulness Ratings

Daily Token Trend

Size vs Usefulness

Warning

Firewall blocked 1 domain

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

• api.anthropic.com

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

network:
  allowed:
    - defaults
    - "api.anthropic.com"

See Network Configuration for more information.

Generated by 🔍 GitHub MCP Structural Analysis · claude · agent · 154.3 AIC · ⌖ 9.09 AIC · ⊞ 11.7K · ◷

• expires on Sep 24, 2026, 4:05 AM UTC-08:00

[github-actions[bot]]

This discussion has been marked as outdated by GitHub MCP Structural Analysis.

A newer discussion is available at Discussion #63172.