[mcp-analysis] MCP Structural Analysis - 2026-05-19

[github-actions[bot]]

Overview

Analyzed 10 GitHub MCP tools across 9 toolsets on 2026-05-19. Total response size: 17,720 tokens (up from 14,250 on 2026-05-18, +24% — driven mostly by larger list_issues body and growing list_label count). Average usefulness for agentic work holds steady at 3.6/5.

• ✅ Best rated: list_pull_requests and search_repositories — 5/5
• ⚠️ Worst rated: get_me — 1/5 (third consecutive 403 failure in workflow context)
• 🔥 Biggest context risk: get_file_contents (8K tokens, no offset/limit), list_label (3.5K, ignores per_page), list_workflows (3.5K, ignores per_page=1)

Critical Issues

• get_me consistently returns 403 Resource not accessible by integration in the GitHub Actions integration context — agents needing actor identity must use the workflow actor env var instead.
• list_workflows and list_label silently ignore per_page=1 — returning 30 and ~110 entries respectively. This is a recurring context-bloat regression for two days running.

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	10
Toolsets Tested	9
Total Tokens (Today)	17,720
Avg Usefulness Rating	3.6/5
Best Rated Tool	list_pull_requests, search_repositories (5/5)
Worst Rated Tool	get_me (1/5)

Usefulness Ratings for Agentic Work

Tool	Toolset	Rating	Assessment
list_pull_requests	pull_requests	⭐⭐⭐⭐⭐	Rich nested head/base/repo, user, draft/merged flags — immediately actionable
search_repositories	search	⭐⭐⭐⭐⭐	minimal_output default + rich metadata (stars, topics, language)
get_file_contents	repos	⭐⭐⭐⭐	Complete file content with SHA/mimeType/URI but no offset/limit
list_issues	issues	⭐⭐⭐⭐	Respects perPage; full body included can balloon response
list_code_scanning_alerts	code_security	⭐⭐⭐⭐	Lean empty-array on no matches; requires state+severity guards
list_discussions	discussions	⭐⭐⭐⭐	Cursor-based pageInfo + totalCount; clean
list_branches	repos	⭐⭐⭐⭐	Concise; respects perPage; protected flag actionable
list_workflows	actions	⭐⭐⭐	Ignores per_page=1; verbose node_id/badge_url per row
list_label	labels	⭐⭐	No per_page parameter; floods context on label-heavy repos
get_me	context	⭐	403 in workflow integration context — unusable

Schema Analysis

Tool	Type	Depth	Key Fields	Notes
get_me	error	1	error	403 wrapper
get_file_contents	object	3	content, resource, mimeType, text, uri	Resource envelope with text payload
list_issues	array	3	number, title, state, body, labels, user, dates	Nested user/labels
list_pull_requests	array	4	number, title, head, base, user, labels, draft, merged	Deeply nested head/base/repo
list_workflows	object	2	total_count, workflows[id,name,path,state,urls]	Wrapper + array
list_code_scanning_alerts	array	1	(empty)	Flat array
list_discussions	object	2	discussions, pageInfo, totalCount, category	GraphQL-style with cursor
list_label	object	2	labels, color, description, id, name, totalCount	Bulk return without paging
search_repositories	array	1	name, full_name, stars, forks, language, topics	Flat with minimal_output
list_branches	array	1	name, sha, protected	Flat and lean

Response Size Analysis (Today)

Toolset	Avg Tokens	Tool(s)
repos	4,020	get_file_contents, list_branches
actions	3,500	list_workflows
labels	3,500	list_label
issues	1,600	list_issues
pull_requests	750	list_pull_requests
discussions	150	list_discussions
search	140	search_repositories
context	25	get_me
code_security	15	list_code_scanning_alerts

Tool-by-Tool Analysis

Tool	Toolset	Tokens	Schema	Rating	Notes
get_me	context	25	error/1	1	403 Resource not accessible by integration
get_file_contents	repos	8,000	object/3	4	Full README.md (~32KB); no offset/limit
list_issues	issues	1,600	array/3	4	perPage respected; long bodies inflate response
list_pull_requests	pull_requests	750	array/4	5	Rich nested metadata; perPage respected
list_workflows	actions	3,500	object/2	3	Ignores per_page=1 (returned 30 of 298)
list_code_scanning_alerts	code_security	15	array/1	4	Empty array — lean
list_discussions	discussions	150	object/2	4	Cursor + totalCount (6,298)
list_label	labels	3,500	object/2	2	No per_page; floods context
search_repositories	search	140	array/1	5	minimal_output + rich metadata
list_branches	repos	40	array/1	4	Concise; protected flag exposed

30-Day Trend Summary

Metric	Value
Data Points	20 across 2 days
Avg Daily Tokens	15,985
Trend (day-over-day)	+24% tokens; ratings flat at 3.6/5
Status	Stable structure; growth driven by list_issues body + list_label count

Recommendations

• Prefer for agents (high rating, low tokens): search_repositories, list_branches, list_discussions, list_pull_requests
• Use with caution (high context cost): get_file_contents (chunk large files manually), list_label (filter post-hoc), list_workflows (cache results)
• Avoid in workflow context: get_me — substitute $GITHUB_ACTOR env var
• Server improvements worth filing: add per_page support to list_label; honor per_page in list_workflows; offset/limit on get_file_contents

Visualizations

Response Size by Toolset

Usefulness Ratings

Daily Token Trend

Size vs Usefulness

References:

• §26098274961

Generated by 🔍 GitHub MCP Structural Analysis · ● 16.6M · ◷

• expires on May 20, 2026, 12:58 PM UTC

[github-actions[bot]]

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

A newer discussion is available at Discussion #33539.