A Model Context Protocol (MCP) server that provides AI-powered Google search using Gemini with search grounding
The ask_google tool is well-defined with explicit schema, good descriptions, and strong tool annotations. The tool exhibits production-grade quality: comprehensive input schema with dual-parameter alias support (question/query), extensive output schema documenting grounding status and citations, parameter descriptions that guide LLM behavior (e.g., 'Prefer current/latest/as of today'), and security/idempotency annotations. Security measures are evident in source URL validation and markdown escaping. However, the single-tool design limits composition opportunities, and the server is STDIO-only which impacts protocol accessibility.
Gemini with Google Search grounding. Use for current/latest facts that post-date your training: versions, releases, API changes, changelogs, breaking news, on-demand web research. Do not use for stable syntax or knowledge already in your training. Short lookups or multi-paragraph briefs both work.
STDIO transport only, not remotely accessible and cannot be used with hosted MCP clients
Single tool limits composition, no batch variants or complementary tools for related research tasks
Output schema includes diagnostics (model, attempts, ttft_ms) that may not be necessary for agent reasoning and dilute response tokens
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 64 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 11 | - | v1 |