[mcp-analysis] MCP Structural Analysis - January 21, 2026

[github-actions[bot]]

Today's analysis tested 13 representative tools across 10 GitHub MCP toolsets, measuring response sizes and rating structural usefulness for agentic workflows. Key findings: 5 excellent tools (rating 5/5), critical bloat issues in list_releases (210KB for 1 item) and list_code_scanning_alerts (97KB for 4 items), and strong performance from core repository operations.

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	13
Total Tokens (Today)	95,200
Average Usefulness Rating	3.92/5
Best Rated Tool	get_label: 5/5
Worst Rated Tool	list_releases: 2/5
Data Points (30-day)	141

Usefulness Ratings for Agentic Work

⭐⭐⭐⭐⭐ Excellent (Rating 5/5) - 7 tools

Tool	Toolset	Tokens	Assessment
get_label	labels	30	Perfect minimal response with flat structure - immediately actionable, no bloat
list_branches	repos	50	Perfect minimal branch listing with flat structure - essential data only
list_tags	repos	90	Excellent minimal tag data with commit reference and download URLs
list_workflows	actions	280	Clean efficient structure with exactly the needed workflow metadata
list_discussions	discussions	320	Well-structured with pagination and category info - efficient token usage
list_issues	issues	620	Excellent structure with complete issue data, pagination, embedded user/label details
get_file_contents	repos	1,600	Perfect file content delivery as resource with clean metadata separation

⭐⭐⭐⭐ Good (Rating 4/5) - 2 tools

Tool	Toolset	Tokens	Assessment
search_repositories	search	350	Efficient search with minimal repo metadata - good for discovery
list_commits	repos	420	Good commit history with nested author/committer details - reasonable balance

⭐⭐⭐ Adequate (Rating 3/5) - 1 tool

Tool	Toolset	Tokens	Assessment
list_pull_requests	pull_requests	14,700	Very comprehensive but includes redundant nested repo objects in head/base - 59KB for perPage=1

⭐⭐ Limited (Rating 2/5) - 2 tools

Tool	Toolset	Tokens	Assessment
list_code_scanning_alerts	code_security	24,200	SEVERE BLOAT - 97KB for 4 alerts, includes full HTML help text - needs summary mode
list_releases	releases	52,500	CATASTROPHIC BLOAT - 210KB for perPage=1! Massive release body + 15 assets with full uploader objects

⭐ Poor (Rating 1/5) - 1 tool

Tool	Toolset	Tokens	Assessment
get_me	context	40	403 permission error - not accessible by integration, completely unusable

Schema Analysis

Tool	Type	Depth	Key Fields	Notes
get_me	error	1	error_message	Permission error - unusable
get_file_contents	resource	2	text, mimeType, uri	Clean resource delivery
list_issues	object	4	issues, pageInfo, totalCount, labels, user	Rich nested structure
list_pull_requests	array	5	number, title, state, head, base	Deeply nested with full repos
list_workflows	object	2	total_count, workflows, id, name	Simple, efficient
list_code_scanning_alerts	array	4	number, rule, state, tool	Bloated with HTML content
get_label	object	1	name, color, description, id	Flat and perfect
list_discussions	object	3	discussions, category, pageInfo	Well-organized GraphQL
search_repositories	object	2	total_count, items	Clean search results
list_releases	array	5	tag_name, body, assets, author	Massive nested assets
list_branches	array	1	name, sha, protected	Flat and minimal
list_commits	array	3	sha, commit, author, committer	Nested author details
list_tags	array	2	name, commit, zipball_url	Simple with references

Response Size Analysis

Toolset	Avg Tokens	Tools Tested	Category
releases	52,500	1	🔴 Catastrophic bloat
code_security	24,200	1	🔴 Severe bloat
pull_requests	14,700	1	🟠 Moderate bloat
repos	552	4	🟢 Efficient
issues	620	1	🟢 Efficient
search	350	1	🟢 Efficient
discussions	320	1	🟢 Efficient
actions	280	1	🟢 Efficient
labels	30	1	🟢 Excellent
context	40	1	⚠️ Error state

Tool-by-Tool Analysis

Repos Toolset (Excellent Performance)

• get_file_contents (1,600 tokens, 5/5): Perfect for code analysis
• list_branches (50 tokens, 5/5): Minimal and efficient
• list_tags (90 tokens, 5/5): Clean tag metadata
• list_commits (420 tokens, 4/5): Good balance of data

Issues & PRs (Mixed Performance)

• list_issues (620 tokens, 5/5): Excellent structure with pagination
• list_pull_requests (14,700 tokens, 3/5): Bloated with nested repo objects

Actions & Workflows (Excellent)

• list_workflows (280 tokens, 5/5): Perfect balance of metadata

Security Tools (Critical Bloat Issue)

• list_code_scanning_alerts (24,200 tokens, 2/5): 97KB for 4 alerts - needs summary option

Releases (Critical Bloat Issue)

• list_releases (52,500 tokens, 2/5): 210KB for 1 release - unusable for agents with context limits

Labels, Discussions, Search (Excellent)

• get_label (30 tokens, 5/5): Perfect minimal response
• list_discussions (320 tokens, 5/5): Well-structured
• search_repositories (350 tokens, 4/5): Efficient discovery

30-Day Trend Summary

Metric	Value
Data Points	141 measurements
Date Range	Jan 5 - Jan 21, 2026
Days Tracked	11 days
Average Daily Tokens	~8,654 tokens
Trend	Stable with occasional spikes from bloated tools

Key Observations:

• Core repository operations (branches, tags, commits) consistently efficient (50-420 tokens)
• Pull requests showing high variance (7,800 - 69,503 tokens) depending on PR complexity
• Security scanning alerts extremely bloated across all measurements (19,484 - 25,000 tokens)
• Releases showing catastrophic bloat (29,300 - 282,700 tokens) based on asset counts

Recommendations

🟢 High-Value Tools (Rating 4-5, Low Tokens)

• get_label - 30 tokens, perfect structure
• list_branches - 50 tokens, essential for git ops
• list_tags - 90 tokens, clean tag data
• list_workflows - 280 tokens, excellent for CI/CD
• list_discussions - 320 tokens, community engagement
• search_repositories - 350 tokens, efficient discovery
• list_commits - 420 tokens, good history tracking
• list_issues - 620 tokens, rich issue data
• get_file_contents - 1,600 tokens, essential for code reading

🟠 Tools Needing Improvement (Rating 2-3)

• list_pull_requests (14,700 tokens, rating 3/5): Consider flattening nested repo objects in head/base
• list_code_scanning_alerts (24,200 tokens, rating 2/5): Add summary mode without full HTML help text
• list_releases (52,500 tokens, rating 2/5): Critical - add options to exclude assets or body content

🔴 Critical Issues

1. list_releases: 210KB for single release with 15 assets - agents hit context limits quickly
2. list_code_scanning_alerts: 97KB for 4 alerts due to verbose HTML descriptions
3. get_me: 403 permission error - tool is completely unusable

💡 Context-Efficient Tool Selection
For agents with context constraints:

• Use list_branches, list_tags, get_label for lightweight operations
• Use list_issues and list_workflows for moderate data needs
• Avoid list_releases and list_code_scanning_alerts without filtering options
• Consider list_pull_requests only when full PR context is essential

🎯 Recommended Tool Enhancements

1. Add exclude_assets=true parameter to list_releases
2. Add summary_only=true parameter to list_code_scanning_alerts
3. Add minimal=true parameter to list_pull_requests to exclude nested repo objects
4. Fix permissions for get_me or document its limitations

Analysis Methodology

This analysis systematically tests representative tools from each GitHub MCP toolset using minimal parameters (perPage=1 where applicable) to measure baseline response sizes. Each tool's response is analyzed for:

1. Token count (1 token ≈ 4 characters)
2. Schema structure (data type, nesting depth, key fields)
3. Usefulness rating (1-5 scale based on completeness, actionability, clarity, efficiency)

Data is stored in cache-memory and tracked over a 30-day rolling window to identify trends and patterns.

---

Data Source: 141 measurements across 11 days (Jan 5-21, 2026)
Workflow Run: §21207220526

AI generated by GitHub MCP Structural Analysis

• expires on Jan 28, 2026, 11:11 AM UTC