Loading ToolBench

Reading the current corpus…

langsearch-mcp-server MCP server · C · ToolBench

langsearch-mcp-server

github.com/fusman60/langsearch-mcp-server

MCP server for LangSearch Web Search & Semantic Rerank API. Use with Claude Code, Claude Desktop, or any MCP-compatible client.

2025-06-18+STDIO2 tools

63/100

Definition Quality88

Protocol Readiness50

Supportability31

Spec posture

Inferred effective spec 2025-06-18+·current baseline 2026-07-28

Static source inference · medium confidence · evidence: structured output

No deprecated protocol patterns detected

Summary

Two tools with complete input schemas and detailed descriptions. Both tools have proper parameter typing, enums, and constraints. Descriptions are comprehensive (200+ chars) and include error handling guidance. Tool names follow verb_noun pattern (langsearch_web_search, langsearch_semantic_rerank). Output schemas are documented in code (formatSearchMarkdown/Json, formatRerankMarkdown/Json). Main gaps: no output schema formally declared in tool registration; no tool annotations (readOnlyHint/destructiveHint); error handling is server-side only, not surfaced to LLM via structured error responses.

Tools (2)

langsearch_semantic_rerankread onlyauthsource verified85/100

Rerank documents by semantic relevance using LangSearch neural reranker. Args: - query: query to rank against - documents: array of document texts (1-100) - top_n: return top N results (optional) - response_format: markdown or json Examples: - Rerank search results for "transformer attention mechanisms" - RAG pipeline document triage Error Handling: - 401/403: Invalid API key - 429: Rate limit - 400: Invalid model or document count

langsearch_web_searchread onlyauthsource verified90/100

Search the web using LangSearch API. Returns titles, URLs, snippets, and optional summaries. Supports freshness filters and configurable result counts. Args: - query: search query (1-500 chars) - count: results 1-10 (default 10) - freshness: oneDay|oneWeek|oneMonth|oneYear|noLimit - summary: include text summaries - response_format: markdown or json Examples: - "latest AI regulation news" → freshness="oneWeek" - "Claude Code documentation" → default settings Error Handling: - 401/403: Invalid API key - 429: Rate limit - 400: Invalid parameters

Issues (5)

HIGHtool ↗

Output schemas not formally declared in tool registration. Formatters (formatSearchMarkdown, formatSearchJson, formatRerankMarkdown, formatRerankJson) exist in code but are not exposed as part of the tool definition. LLMs cannot see what fields to expect in responses.

langsearch_web_searchlangsearch_semantic_rerank

MEDIUMtool-annotations ↗

Tool annotations missing. Neither tool declares readOnlyHint (both are read-only), destructiveHint, or idempotentHint. This prevents clients from optimizing caching, retry logic, and safety checks.

langsearch_web_searchlangsearch_semantic_rerank

MEDIUMrecovery-guide ↗

Error handling is server-side only. handleApiError() function returns user-friendly messages but these are not structured or categorized (retryable vs user-fixable vs fatal). LLMs receive plain text errors without recovery guidance.

langsearch_web_searchlangsearch_semantic_rerank

Recommendations

Add formal output schema declarations to both tools. Define return type as object with fields: query, total, count, results (array of {title, url, snippet, summary, datePublished}). Include in tool registration via outputSchema property.
Add tool annotations: set readOnlyHint=true for both tools since they only read data and have no side effects. This enables client-side caching and retry optimization.
Restructure error responses as objects with {code, message, retryable, userFixable, suggestion} to guide LLM recovery logic. Example: {code: 'RATE_LIMIT', message: 'Rate limit exceeded', retryable: true, suggestion: 'Wait 60 seconds before retrying'}.
Document API key setup in README with link to https://langsearch.com/dashboard and example: LANGSEARCH_API_KEY=sk_... node src/index.js. Add validation message: 'LANGSEARCH_API_KEY not set. Get one at https://langsearch.com/dashboard'.
Add optional top_n parameter to langsearch_web_search (default 10, max 10) to match langsearch_semantic_rerank pattern. Document that results are ordered by relevance.

Score history

Overall score trend

First recorded score · v2 rubric

63/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	C	63	2025-06-18+	v2