[mcp-analysis] MCP Structural Analysis - 2026-08-18

[github-actions[bot]]

Analyzed 10 representative GitHub MCP tools (one per toolset) for response size, schema structure, and usefulness for agentic work. Today's total footprint was 6,268 tokens across all calls, average usefulness 3.6/5. Standouts: list_pull_requests, list_workflows (mcpscripts wrapper), and the injected <github-context> block all scored 5/5. Weakest: get_teams, blocked outright by a sandbox permission gate. This is the first data point — no historical trend yet (30-day window starts today).

Key Findings

• The mcpscripts wrappers (list_workflows, get_label) beat their native GitHub MCP equivalents on structure: flat JSON, pagination metadata, zero icon/meta overhead.
• Every native GitHub MCP call repeats a ~3KB base64 icon pair in _meta.serverInfo, regardless of payload size — pure fixed overhead that adds up across a session.
• list_discussions and search_code nest their real payload as an escaped JSON string inside content[].text, forcing an extra parse step before the data is usable.
• fields-filtered calls (list_issues, list_pull_requests) were the leanest and most directly actionable of all native tools tested.
• get_teams failed at the permission layer before returning any data — a hard blocker for unattended agent runs.

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	10
Total Tokens (Today)	6,268
Average Usefulness Rating	3.6/5
Best Rated Tools	github_context, list_pull_requests, list_workflows: 5/5
Worst Rated Tool	get_teams: 1/5

Usefulness Ratings for Agentic Work

Tool	Toolset	Rating	Assessment
github_context	workflow_context	⭐⭐⭐⭐⭐	Excellent — zero-cost identity block, scopes other calls
list_pull_requests	pull_requests	⭐⭐⭐⭐⭐	Excellent — clean, flat, fields-filtered
list_workflows (mcpscripts)	actions	⭐⭐⭐⭐⭐	Excellent — lean JSON, real pagination, no bloat
list_issues	issues	⭐⭐⭐⭐	Good — lean with fields filter, one result withheld by integrity policy
get_label (mcpscripts)	labels	⭐⭐⭐⭐	Good — clean wrapper output
get_file_contents	repos	⭐⭐⭐	Adequate — usable content, heavy fixed icon overhead
list_code_scanning_alerts	code_security	⭐⭐⭐	Adequate — clean shape but empty result this run
list_discussions	discussions	⭐⭐⭐	Adequate — useful pagination, but nested JSON-in-string payload
search_code	search	⭐⭐⭐	Adequate — useful total_count, same nested-string issue
get_teams	context	⭐	Poor — blocked by sandbox permission gate, no data returned

Schema Analysis

Tool	Type	Depth	Key Fields	Notes
github_context	markdown	1	actor, repository, workspace, workflow-run-id	Prompt-injected, no MCP call
get_teams	n/a	0	—	Permission-denied before response
get_file_contents	object	4	content[].resource.text/uri, _meta.serverInfo.icons	Icon blob repeated every call
list_issues	array	1	number, title, state, labels	Fields-filtered, lean
list_pull_requests	array	1	number, title, state, html_url	Fields-filtered, lean
list_workflows	object	2	total_count, workflows[].id/name/path/state	No meta bloat
list_code_scanning_alerts	array	1	—	Empty this run
list_discussions	object	3	discussions[].category/html_url/number/title/user, pageInfo, totalCount	JSON-in-string payload
get_label	object	2	labels[].id/name/color/description	No meta bloat
search_code	object	3	items[].name/path, total_count, incomplete_results	JSON-in-string payload

Response Size Analysis

Toolset	Avg Tokens	Tools Tested
repos	4,050	1
discussions	925	1
search	840	1
actions	172	1
labels	106	1
workflow_context	80	1
issues	45	1
pull_requests	45	1
code_security	5	1
context	0	1 (blocked)

Tool-by-Tool Analysis

Tool	Toolset	Tokens	Schema	Rating	Notes
github_context	workflow_context	80	markdown	5	Zero-cost identity block
get_teams	context	0	n/a	1	Rejected by permission gate
get_file_contents	repos	4,050	object (depth 4)	3	README + repeated icon blob + long contributor list
list_issues	issues	45	array (depth 1)	4	One match withheld by integrity policy, disclosed clearly
list_pull_requests	pull_requests	45	array (depth 1)	5	Clean and actionable
list_workflows	actions	172	object (depth 2)	5	mcpscripts wrapper, no bloat
list_code_scanning_alerts	code_security	5	array (depth 1)	3	Empty result set
list_discussions	discussions	925	object (depth 3)	3	JSON-in-string + icon blob
get_label	labels	106	object (depth 2)	4	mcpscripts wrapper, clean
search_code	search	840	object (depth 3)	3	JSON-in-string + icon blob

30-Day Trend Summary

Metric	Value
Data Points	1 day (first run)
Average Daily Tokens	6,268
Average Rating Trend	Not yet established — baseline

Recommendations

• High-value tools (rating 4–5): github_context, list_pull_requests, list_workflows, list_issues, get_label — prefer these when fields filtering or mcpscripts wrappers are available.
• Tools needing improvement: get_teams (permission-gated in this sandbox), list_discussions and search_code (unwrap payload from content[].text JSON string server-side to save a parse step).
• Context-efficient tools (low tokens, high rating): list_pull_requests, list_issues, list_workflows, get_label — all under 200 tokens with fields filtering.
• Context-heavy tools: get_file_contents (4,050 tokens, dominated by fixed icon metadata and long README content unrelated to the file itself) — consider trimming _meta.serverInfo.icons from responses or offering a raw-content-only mode.

Visualizations

Response Size by Toolset

Usefulness Ratings

Daily Token Trend

Size vs Usefulness

Generated by 🔍 GitHub MCP Structural Analysis · agent · 126.5 AIC · ⌖ 16.4 AIC · ⊞ 12.2K · ◷

• expires on Aug 19, 2026, 4:14 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 #53988.