[mcp-analysis] MCP Structural Analysis - 2026-09-22

[github-actions[bot]]

Overview

Tested 10 representative GitHub MCP tools across 9 toolsets plus the injected workflow-context block. Average usefulness today: 3.4/5. Response sizes ranged from 46 tokens (list_pull_requests) to 36,602 tokens for list_code_scanning_alerts, which again exceeded the MCP 25k-token hard limit and had to be read from a saved file rather than inline.

Best rated: github_context / list_workflows (5/5) — zero-cost identity block and the clean mcpscripts wrapper.
Worst rated: list_issues / list_code_scanning_alerts (2/5) — the former blocked by an integrity-policy redaction that recurs every single day (16/16); the latter now a stable steady-state condition, not a one-off spike (matches the 2026-09-15 measurement almost exactly: same 3 rule IDs, same ~30-alert scale).

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	10
Total Tokens (Today)	41,346
Average Usefulness Rating	3.4/5
Best Rated Tool	github_context / list_workflows: 5/5
Worst Rated Tool	list_issues / list_code_scanning_alerts: 2/5
Historical Data Points (30-day window)	160 across 16 measured days

Usefulness Ratings for Agentic Work (Today)

Tool	Toolset	Rating	Assessment
github_context	workflow_context	⭐⭐⭐⭐⭐	Excellent — zero-cost, immediately actionable identity
list_workflows (mcpscripts)	actions	⭐⭐⭐⭐⭐	Excellent — clean JSON, explicit pagination, no bloat
search_code	search	⭐⭐⭐⭐	Good — embedded code fragment avoids follow-up read
list_pull_requests	pull_requests	⭐⭐⭐⭐	Good — flat, minimal, directly actionable
get_file_contents	repos	⭐⭐⭐	Adequate — usable, but icon/_meta overhead persists
list_discussions	discussions	⭐⭐⭐	Adequate — good pagination metadata, fixed overhead
get_label	labels	⭐⭐⭐	Adequate — tiny payload swamped by _meta block
search_users	users	⭐⭐⭐	Adequate — minimal identity fields only
list_code_scanning_alerts	code_security	⭐⭐	Limited — complete but extremely token-inefficient
list_issues	issues	⭐⭐	Limited — redacted item blocks assessment

Schema Analysis

Tool	Type	Depth	Key Fields	Notes
github_context	markdown	1	actor, repository, workspace, workflow-run-id	Prompt-injected, no MCP call
list_workflows	object	2	total_count, workflows[].id/name/path/state	Clean wrapper schema
search_code	object	2	items[].name/path, text_matches[].fragment	Fragment reduces follow-up calls
list_pull_requests	array	1	number, title, state, html_url	Flat, field-filtered
get_file_contents	object	3	content.resource.text, _meta.serverInfo	Constant 2235-char icon overhead
list_discussions	object	3	discussions[].category/user, pageInfo, totalCount	Nested relationships present
get_label	object	1	name, color, description, id	Tiny payload, same icon overhead
search_users	object	2	items[].login/id/profile_url	No name/bio/role fields
list_code_scanning_alerts	array	3	rule.id/severity/full_description/help	Verbatim-repeated rule text per instance
list_issues	array	2	number, title, state, labels	One item redacted by integrity policy

Response Size Analysis

Toolset	Avg Tokens (30-day)	Tools Tested
code_security	6,889	16
repos	2,581	16
labels	847	16
discussions	694	16
users	618	16
search	608	16
actions	171	16
issues	96	16
pull_requests	93	16
workflow_context	92	16

Tool-by-Tool Analysis (Today)

Tool	Toolset	Tokens	Schema	Rating	Notes
github_context	workflow_context	90	markdown	5	Zero-cost identity block
list_workflows	actions	172	object	5	687-char clean JSON via mcpscripts
list_pull_requests	pull_requests	46	array	4	Single Dependabot PR, flat
search_code	search	726	object	4	Fragment embedded, no rate-limit today
get_label	labels	597	object	3	~25 tokens of real content, rest is icon overhead
search_users	users	621	object	3	Identity only, no profile detail
list_discussions	discussions	707	object	3	totalCount 10,821 (up from 10,698 on 09-15)
get_file_contents	repos	1,720	object	3	.gitignore grew again (6,896 chars total)
list_issues	issues	65	array	2	Empty + 1 item redacted by integrity policy
list_code_scanning_alerts	code_security	36,602	array	2	30 alerts, 3 unique rule IDs, exceeded 25k limit

30-Day Trend Summary

Metric	Value
Data Points	160
Measured Days	16 (2026-08-25 → 2026-09-22)
Average Daily Total Tokens	12,688
Average Rating Trend	Stable — code_security and issues chronically low; actions/workflow_context/pull_requests chronically high

Recommendations

• High-value tools (rating 4-5): github_context (free), list_workflows via the mcpscripts wrapper, list_pull_requests with field selection, search_code (when not rate-limited).
• Tools needing improvement: list_code_scanning_alerts — needs rule-level deduplication instead of repeating full rule.full_description/help text per instance (would cut ~36.6k tokens dramatically); list_issues — integrity-policy redaction has blocked one item on all 16 measured days and should be revisited.
• Context-efficient tools (low tokens, high rating): list_workflows, list_pull_requests, github_context.
• Context-heavy tools: list_code_scanning_alerts (6,889 avg, up to 36.6k), get_file_contents when targeting growing files like README.md (peaked at 11,900 tokens on 2026-09-14).

Visualizations

Response Size by Toolset

Usefulness Ratings

Daily Token Trend

Size vs Usefulness

References: §35724109792

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 · 170.5 AIC · ⌖ 8.14 AIC · ⊞ 11.7K · ◷

• expires on Sep 23, 2026, 4:04 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 #62941.