MCP server providing full-text search and documentation access for NeuroLink AI toolkit
NeuroLink MCP server presents a documentation-focused toolkit with 6 READ_ONLY tools. Tool naming follows verb_noun convention (search_docs, get_page, list_sections, get_api_reference, get_examples, get_changelog). All tools have descriptions between 80-150 chars, meeting the 10-1024 char guideline. Input schemas are present for all tools with typed parameters. However, there are notable gaps: output schemas are not documented in the provided code, parameter descriptions lack depth and formatting constraints, and error handling patterns are not visible. The server operates on STDIO transport only, which caps protocol readiness at 50 and indicates this is a local/CLI integration rather than production agent-ready. All tools are read-only with low security risk, but the lack of output documentation and LLM-optimized descriptions limits agent reasoning quality.
Get SDK API reference documentation. Optionally filter by method name.
Get recent changelog entries for NeuroLink releases.
Get code examples from the documentation. Optionally filter by topic or provider.
Get the full content of a specific documentation page by its path.
List all documentation sections and their pages.
Full-text search across all NeuroLink documentation. Returns ranked results with title, description, section, and path.
Output schemas not documented. LLMs cannot infer the structure of returned data (fields, types, pagination). This forces agents to guess at response formats and complicates chaining with other tools.
Parameter descriptions lack formatting constraints and validation rules. Example: 'search_docs' limit param says 'Max results to return (default: 10, max: 50)' but does not state minimum or whether zero is valid. Query param description does not specify character length limits or unsupported characters.
No error handling patterns visible. Tools lack descriptions of failure modes, recovery guidance, or error messages. If 'get_page' receives an invalid path, will it return 404 with suggestions? Will it say 'Path not found. Available sections: ...'? Unknown.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 24 | 2026-07-28+ | v2 |
| 2026-03-09 | B | 70 | 1.26.0+ | v1 |
Parameters accept string enums (section: 'features|sdk|cli|mcp|rag|examples'; provider: 'openai|anthropic|google') but these are mentioned only in descriptions, not as JSON Schema enum constraints. LLMs cannot parse enum values from natural language descriptions reliably.
Tool descriptions are under 200 chars but lack LLM-optimization. They state WHAT but not WHY or WHEN. Example: 'search_docs' says 'Full-text search across all NeuroLink documentation' but does not say 'Use this first to find relevant docs before calling get_page for full content.' Missing dependency hints and disambiguation vs similar tools.
No pagination guidance. If 'search_docs' returns 50+ results, LLMs cannot know whether the response is capped or complete. If pagination is supported, the output schema must include total_count or next_cursor. If not supported, description should state max limit enforced.