Loading ToolBench

Reading the current corpus…

utility-mcp-server MCP server · B · ToolBench

utility-mcp-server

github.com/realmcpservers/utility-mcp-server

A utility Model Context Protocol (MCP) server with remote repository support for release notes generation and git operations

2026-07-28+STREAMABLE_HTTP1 tools

75/100

Definition Quality81

Protocol Readiness59

SupportabilityNot measured-

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence · detected: Logging, Dynamic Client Registration

Deprecated protocol patterns detected

Summary

Single-tool server with a well-structured release notes generator. Tool naming follows verb_noun convention ('generate_release_notes'), description is comprehensive (220 chars, within 10-1024 ideal range), and input schema is fully typed with descriptions. However, output schema is not documented in the provided source, and no error handling guidance is visible. Parameter descriptions are clear and specific (e.g., 'Release date in YYYY-MM-DD format'), covering format constraints. The tool accepts multiple identifier patterns (repo_path for local, repo_url for remote) and auth tokens, reducing friction. No tool annotations present (readOnlyHint, destructiveHint, idempotentHint) despite the risk classification as 'READ_ONLY'. Security concern: GitHub and GitLab tokens are exposed as optional parameters, should use server-side secret injection instead.

Tools (1)

generate_release_notesread onlysource verified81/100

Generate release notes from git commits between tags. Fetches git commits between two tags from local or remote repositories (GitHub, GitLab) and returns structured data for AI-powered categorization and release notes generation. The tool automatically detects the previous tag if not provided, validates that tags exist, and extracts PR/MR numbers from commits.

Issues (5)

HIGHresponse-shaper ↗

Output schema not documented. LLMs cannot infer what fields the tool returns, forcing them to reason about downstream data extraction and blocking tool chaining.

generate_release_notes

HIGHsecret-injection ↗

GitHub and GitLab tokens exposed as optional input parameters. Secrets in parameters leak into agent traces and logs. Should use server-side secret injection via environment variables or vault.

generate_release_notes

MEDIUMtool-annotations ↗

No tool annotations present (readOnlyHint, destructiveHint, idempotentHint). Tool is marked READ_ONLY in metadata but schema lacks structured annotation. Should add 'readOnly': true to descriptor.

generate_release_notes

MEDIUMrecovery-guide ↗

No error handling guidance visible in source. Tool should return actionable error messages. E.g., if tags don't exist: 'Tag v0.50.0 not found. Available tags: v0.49.0, v0.48.5, ...' instead of raw API errors.

generate_release_notes

Recommendations

Document the output schema explicitly. E.g., 'Returns object { commits: [ { message, author, hash, pr_number?, ... } ], tag_range: string, total_count: number, statistics: { breaking: int, features: int, bugfixes: int, ... } }'.
Move github_token and gitlab_token from optional input parameters to server-side environment variables (GITHUB_TOKEN, GITLAB_TOKEN). Update schema to remove these params and add a note: 'Tokens are injected server-side; ensure they are set in your environment.'
Add tool annotation 'readOnly': true to the fastMCP tool descriptor to signal to clients that this tool does not modify state.
Enhance error handling with specific guidance. If tag lookup fails, catch the error and return: 'error': 'Tag <tag> not found in <repo>. Available recent tags: [list]. Verify the tag name and try again.' This enables LLM self-correction.
Add parameter constraint documentation for 'release_date': state that the format must be YYYY-MM-DD (ISO 8601) and is optional; if omitted, defaults to today's date. Clarify in description: 'Optional release date in YYYY-MM-DD format. If omitted, defaults to today.'
Test pagination scenarios: if a tag range has hundreds of commits, document the limit (e.g., 'Returns up to 500 commits per call') and provide a cursor or offset mechanism.
Add to description: 'This tool is read-only and does not modify repositories. It requires at least read access to the target repository (public repos require no auth; private repos require a valid GitHub/GitLab token).'

Spec posture evidence

Inferred effective spec: 2026-07-28+.

Relies on Logging (deprecated) - log to stderr or use OpenTelemetry

Score history

Overall score trend

↑ 53 points across a rubric change (v1 → v2)

75/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-22	B	75	2026-07-28+	v2
2026-03-09	F	22	-	v1