Loading ToolBench
Viraill Model Context Protocol (MCP) Server - Generative Engine Optimization (GEO) & Agentic Commerce Readiness
Three tools with complete input schemas and descriptions, but significant quality gaps. Tool names lack action verbs (geo_audit, geo_social_generate, agentic_scan are noun-heavy). Descriptions are verbose (200+ chars) and domain-specific jargon-heavy, making them harder for LLMs to parse. No output schemas documented. Parameters lack descriptions in geo_social_generate (url, seed_topic have none). Error handling is minimal, API errors are passed through without recovery guidance. No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Schemas are present but incomplete: geo_social_generate lacks required field declarations.
Audit a website's Agentic Commerce Readiness (/100 score across 5 pillars: Crawl/Ingestion, OpenAPI, Schema.org, x402 settlement, 8 distribution channels) and generate remediation code.
Audit website visibility in generative search engines (ChatGPT, Perplexity, Gemini, Claude). Full-site crawl (up to 25 pages, bilingual FR/EN matching). Returns score /100 calibrated on a 67-domain reference batch, per-intent coverage gaps, competitive white spaces, conquest plan with projected score (§3bis), RAG-first answer blocks (BLUF) + Schema.org FAQPage markup. Response includes an integrity field (7 anti-hallucination checks) — treat integrity.ok=false as a veto and use the fallback report.
Generate viral social publications (LinkedIn, X/Threads, Debate, Carousel) mathematically aligned with market intent centroids to build external RAG citation consensus.
Tool names lack action verbs. 'geo_audit', 'geo_social_generate', 'agentic_scan' are noun-heavy and do not clearly signal what action the LLM should expect. Should be 'audit_geo_visibility', 'generate_social_content', 'scan_agentic_readiness'.
geo_social_generate parameters 'url' and 'seed_topic' lack descriptions. LLMs cannot infer what 'seed_topic' means (topic for what? how specific?). All parameters must have non-empty descriptions.
No output schemas documented. LLMs do not know what fields to expect from responses (e.g., does geo_audit return 'score', 'integrity', 'coverage_gaps'?). This forces LLMs to guess and parse unstructured JSON.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | 2026-07-28+ | v2 |
Error handling is minimal. API errors (402, 4xx, 5xx) are passed through as raw text without recovery guidance. LLMs receive 'Viraill API Error 402: ...' with no hint about what to do next (retry? ask user? check quota?).
Descriptions are verbose (200+ chars) and use domain jargon ('BLUF', 'RAG-first', 'Schema.org FAQPage', 'x402 settlement') without explaining what these mean to an LLM. Baseline for A+ tools is 50-200 chars with clear, jargon-free language.