[mcp-analysis] MCP Structural Analysis - January 30, 2026

[github-actions[bot]]

Analyzed 12 GitHub MCP tools across 9 toolsets for response size, schema structure, and usefulness for agentic workflows. Key finding: code_security tools show severe bloat (24K tokens for 5 alerts) while repos tools excel at efficiency.

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	12
Toolsets Covered	9
Total Tokens (Today)	50,660
Average Usefulness Rating	4.00/5
Best Rated Tool	get_file_contents: ⭐⭐⭐⭐⭐
Worst Rated Tool	get_me: ⭐ (403 error)

Usefulness Ratings for Agentic Work

Excellent Tools (⭐⭐⭐⭐⭐ Rating 5/5)

Tool	Toolset	Tokens	Assessment
get_file_contents	repos	150	Simple success message with SHA - actual content downloaded, perfect for file operations
list_workflows	actions	850	Clean workflow list with essential metadata - excellent for CI/CD operations
get_label	labels	30	Perfect minimal label response - flat structure, immediately actionable
list_discussions	discussions	280	Clean discussion listing with category and pagination - efficient structure
list_branches	repos	50	Perfect minimal branch listing - essential data only
list_commits	repos	420	Good commit history with author details - reasonable balance
list_tags	repos	90	Excellent minimal tag data with commit reference and URLs

Good Tools (⭐⭐⭐⭐ Rating 4/5)

Tool	Toolset	Tokens	Assessment
list_issues	issues	4,500	Rich issue data with large body text (team status) - generally excellent for tracking work with pagination support
search_repositories	search	290	Efficient search with minimal repo metadata - good for discovery

Adequate Tools (⭐⭐⭐ Rating 3/5)

Tool	Toolset	Tokens	Assessment
list_pull_requests	pull_requests	20,000	Very verbose with deeply nested repo objects in head/base, large PR body text - high token cost but comprehensive data

Tools Needing Improvement (⭐⭐ Rating 2/5)

Tool	Toolset	Tokens	Assessment
list_code_scanning_alerts	code_security	24,000	SEVERE BLOAT - 96KB for 5 alerts, includes full HTML help text and verbose descriptions - needs summary mode

Unusable Tools (⭐ Rating 1/5)

Tool	Toolset	Tokens	Assessment
get_me	context	0	403 permission error - not accessible by integration token

Schema Analysis

Tool	Type	Depth	Key Fields	Structural Notes
get_file_contents	string	1	content, sha	Simple success message pattern
list_issues	object	3	issues, pageInfo, totalCount	GraphQL-style pagination with nested user/label objects
list_pull_requests	array	5	number, title, state, user, head, base	Deeply nested with full repo objects duplicated in head/base branches
list_workflows	object	2	total_count, workflows	Clean REST structure with minimal nesting
list_code_scanning_alerts	array	4	number, rule, state, tool, most_recent_instance	Contains full HTML help text in each alert
get_label	object	1	name, color, description, id	Flat structure, no nesting
list_discussions	object	3	discussions, pageInfo, totalCount	GraphQL-style with category metadata
search_repositories	object	2	total_count, items, incomplete_results	Clean search results structure
list_branches	array	1	name, sha, protected	Flat array, no nesting
list_commits	array	3	sha, commit, author, committer, html_url	Nested commit/author objects
list_tags	array	2	name, commit, zipball_url, tarball_url	Lightweight with commit reference

Response Size Analysis

Toolset	Avg Tokens	Tools Tested	Efficiency Rating
repos	178	4	⭐⭐⭐⭐⭐ Excellent
labels	30	1	⭐⭐⭐⭐⭐ Excellent
discussions	280	1	⭐⭐⭐⭐⭐ Excellent
search	290	1	⭐⭐⭐⭐⭐ Excellent
actions	850	1	⭐⭐⭐⭐ Good
issues	4,500	1	⭐⭐⭐ Adequate
pull_requests	20,000	1	⭐⭐ Heavy
code_security	24,000	1	⭐ Severe Bloat
context	0	1	❌ Error

Critical Findings

🚨 Severe Bloat Issues

1. list_code_scanning_alerts (24,000 tokens)

   • 96KB response for just 5 alerts
   • Includes full HTML-formatted help text for each alert
   • Contains verbose rule descriptions and multiple location instances
   • Recommendation: Add summary mode that returns alert IDs/titles only

2. list_pull_requests (20,000 tokens)

   • Full repository objects duplicated in both head and base branches
   • Large PR body text included in list view
   • Deeply nested assignees, reviewers, and labels arrays
   • Recommendation: Provide minimal mode with repo IDs instead of full objects

✅ Efficiency Champions

1. get_label (30 tokens)

   • Flat structure with only essential fields
   • Zero redundancy
   • Perfect for label operations

2. list_branches (50 tokens)

   • Minimal data: name, SHA, protected flag
   • No unnecessary metadata
   • Ideal for git operations

3. list_tags (90 tokens)

   • Lightweight with commit reference
   • Includes download URLs without bloat

Tool-by-Tool Analysis

Tool	Toolset	Tokens	Schema	Rating	Key Observations
get_me	context	0	error	⭐	403 permission error - not accessible by integration token
get_file_contents	repos	150	string	⭐⭐⭐⭐⭐	Simple success message with SHA - actual content downloaded, perfect for file operations
list_issues	issues	4,500	object	⭐⭐⭐⭐	Rich issue data with large body text (team status) - generally excellent for tracking work with pagination support
list_pull_requests	pull_requests	20,000	array	⭐⭐⭐	Very verbose with deeply nested repo objects in head/base, large PR body text - high token cost but comprehensive data
list_workflows	actions	850	object	⭐⭐⭐⭐⭐	Clean workflow list with essential metadata - excellent for CI/CD operations
list_code_scanning_alerts	code_security	24,000	array	⭐⭐	SEVERE BLOAT - 96KB for 5 alerts, includes full HTML help text and verbose descriptions - needs summary mode
get_label	labels	30	object	⭐⭐⭐⭐⭐	Perfect minimal label response - flat structure, immediately actionable
list_discussions	discussions	280	object	⭐⭐⭐⭐⭐	Clean discussion listing with category and pagination - efficient structure
search_repositories	search	290	object	⭐⭐⭐⭐	Efficient search with minimal repo metadata - good for discovery
list_branches	repos	50	array	⭐⭐⭐⭐⭐	Perfect minimal branch listing - essential data only
list_commits	repos	420	array	⭐⭐⭐⭐	Good commit history with author details - reasonable balance
list_tags	repos	90	array	⭐⭐⭐⭐⭐	Excellent minimal tag data with commit reference and URLs

30-Day Trend Summary

Metric	Value
Data Points Collected	227
Average Daily Token Usage	70,675
Date Range	Jan 5 - Jan 30, 2026
Most Analyzed Toolsets	repos, actions, issues, discussions

Trend Observations

• Repos toolset consistently delivers excellent efficiency (50-420 tokens per tool)
• Code security alerts remain problematic across all measurement days
• Pull requests show high variance (2,850-69,503 tokens) depending on PR content
• Labels and branches maintain perfect minimal responses throughout

Recommendations

For Tool Developers

1. Add Summary Modes

   • list_code_scanning_alerts: Return alert IDs and titles only, require separate call for full details
   • list_pull_requests: Add minimal mode with repo reference IDs instead of full objects
   • list_issues: Consider truncating body text in list views

2. Reduce Redundancy

   • Avoid duplicating full repository objects in head/base references
   • Use reference IDs for related entities when full objects aren't needed
   • Consider field selection parameters (e.g., fields=id,title,state)

3. Optimize HTML Content

   • Strip HTML formatting from help text in API responses
   • Provide help text URLs instead of inline HTML
   • Use markdown instead of HTML where possible

For Agent Developers

1. High-value tools (rating 4-5):

   • repos toolset: get_file_contents, list_branches, list_tags, list_commits
   • actions: list_workflows
   • labels: get_label
   • discussions: list_discussions
   • search: search_repositories

2. Use with caution (rating 2-3):

   • list_code_scanning_alerts: Very high token cost, use sparingly
   • list_pull_requests: Consider fetching by ID instead of listing
   • list_issues: Can be heavy with long body text

3. Avoid (rating 1):

   • get_me: Permission errors in workflow contexts

4. Context-efficient tools (low tokens, high rating):

   • get_label (30 tokens, 5/5)
   • list_branches (50 tokens, 5/5)
   • list_tags (90 tokens, 5/5)
   • list_discussions (280 tokens, 5/5)

Visualizations

Response Size by Toolset

Usefulness Ratings

Daily Token Trend

Size vs Usefulness

---

References:

• Workflow Run: §21513762378

AI generated by GitHub MCP Structural Analysis

• expires on Feb 6, 2026, 11:19 AM UTC

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-02-06T11:19:46.144Z.