Loading ToolBench

Reading the current corpus…

localscope MCP server · D · ToolBench

localscope

github.com/parh0m2007/localscope

Local-first MCP code analyst: index your repo privately (embeddings + ripgrep-style search), then ask 'where does X break if I change Y' — zero cloud, zero config.

2026-07-28+STDIO6 tools

52/100

Definition Quality62

Protocol Readiness50

Supportability38

Spec posture

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

Static source inference · medium confidence

Current-spec patterns detected

Summary

Six tools with complete Zod schemas and consistent descriptions. All parameters have types and descriptions. Tool names are action-oriented (index, search, impact, status, references, definition). Descriptions average ~100 chars, meeting baseline. However, descriptions lack WHEN/WHY guidance and dependency hints. No error handling guidance in descriptions. Output schemas are not documented, LLMs cannot predict response structure. No tool annotations (readOnlyHint). Missing composition guidance for multi-step workflows.

Tools (6)

definitionread only50/100

Locate the definition of a symbol in the codebase. Returns the file and line number where the symbol is defined.

impactread only50/100

Analyze the impact of changing a file or symbol. Shows which files and symbols would break if the target is modified.

indexread onlysource verified80/100

Index a repository for code analysis. Scans files, extracts symbols, builds embeddings and dependency graphs.

referencesread only50/100

Find all call sites and usages of a symbol. Returns locations where the symbol is referenced in the codebase.

searchread only50/100

Search the indexed repository using natural language, symbol names, or code fragments. Returns semantic and lexical matches.

statusread only50/100

Check the indexing status of a repository. Returns information about indexed files, symbols, and index freshness.

Issues (5)

HIGHtool ↗

Output schemas not documented. LLMs cannot predict response structure (fields, types, pagination). Descriptions state WHAT tools do but not WHAT they return.

indexsearchimpactstatusreferencesdefinition

HIGHtool-description ↗

Descriptions lack WHEN/WHY guidance and dependency hints. E.g., 'search' does not explain when to call it vs 'definition' or 'references'. No guidance on multi-step workflows.

searchreferencesdefinition

HIGHrecovery-guide ↗

No error handling guidance in tool descriptions. LLMs cannot know what to do if index fails, search returns no results, or impact analysis times out.

indexsearchimpact

MEDIUMtool ↗

Tool annotations missing. No readOnlyHint, destructiveHint, or idempotentHint. All tools are read-only but this is not declared in schema.

Recommendations

Document output schemas for all tools. E.g., 'Returns: {results: [{file: string, line: number, symbol: string, context: string}], total: number, next_cursor?: string}'. Include pagination structure.
Enhance descriptions with WHEN/WHY guidance. E.g., 'search: Use for natural-language queries across the codebase. Call this first to explore; then use definition() to jump to a specific symbol, or references() to see all usages.'
Add error handling guidance to descriptions. E.g., 'index: If indexing fails due to file size, reduce max_files or exclude large directories. Returns error with retry guidance.'
Add tool annotations to schema. Mark all tools with readOnlyHint: true since they do not modify state. This signals to LLMs that these calls are safe to retry.
Clarify 'path' parameter semantics. Document: 'Must match a previously indexed repository root. If unsure, call status() first to list indexed roots. Defaults to current directory if it has been indexed.'

Score history

Overall score trend

First recorded score · v2 rubric

52/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	D	52	2026-07-28+	v2