Model Context Protocol (MCP) Server for SAGE SEO, Entity AEO, and Generative Engine Optimization (GEO)
Server exposes 10 tools with reasonable naming (verb_noun pattern: audit_*, entity_*, etc.) and descriptions present for all tools. However, critical gaps undermine quality: (1) Input schemas lack proper JSON Schema type definitions, parameters are documented in docstrings but not formally declared in the @app.tool() decorator; (2) Output schemas are completely undocumented, tools return JSON strings but no schema is provided to guide LLM parsing; (3) Parameter descriptions are present but often generic (e.g., 'embedding_backend' lacks enum constraint); (4) Error handling returns structured JSON but lacks recovery guidance or retryability classification; (5) No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all tools being read-only. Baseline: 194 chars avg description length, most tools meet this. Baseline: 100% of A+ tools have documented return types, this server has 0%. Per-tool scores average 52.
Audits Pillar 2 (Answer Engine Optimization) for JSON-LD entity graphs, sameAs authority, and direct-answer density.
[Compatibility Alias] Audits a webpage for GEO standards and RAG readiness via SAGE Core.
Performs a full 3-pillar audit (Technical SEO, Entity AEO, and GEO) on raw HTML content via SAGE Core.
Performs a full 3-pillar audit (Technical SEO, Entity AEO, and GEO) on a live URL via SAGE Core.
Evaluates Citation Survival Proxy (CSP), softmax semantic entropy, and direct-answer density for generative engine retrieval.
Analyzes entity graph completeness, primary entity identification, schema depth, and Wikidata/Wikipedia authority anchors.
No formal input schemas in tool registration. Parameters documented in docstrings but not declared via JSON Schema in @app.tool() decorator. LLMs cannot parse parameter types, constraints, or enums from docstrings alone.
Output schemas completely undocumented. Tools return JSON strings but no schema describes the response structure. LLMs cannot plan downstream tool calls or extract fields reliably.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 51 | <=2025-11-25 | v2 |
Generates an optimized llms.txt manifest for AI crawlers using SAGE semantic passage chunking and CSP analysis.
Audits Pillar 3 (Generative Engine Optimization) evaluating semantic chunking, RAG simulation, and Citation Survival Proxy (CSP).
Returns the list of active SAGE engine capabilities, version provenance, and Evidence Taxonomy levels.
Audits Pillar 1 (Technical SEO) including DOM cleanliness, canonicals, robots.txt AI-bot policies, and headers.
embedding_backend parameter accepts free-form strings ('auto', 'fastembed', 'sentence-transformers', 'hashing') but is not declared as an enum. LLMs may hallucinate invalid backend names.
Error responses lack recovery guidance and retryability classification. _format_error() returns structured JSON but does not tell LLM whether to retry, ask user, or abort. Baseline: 100% of A+ tools classify errors as retryable/user-fixable/fatal.
No tool annotations despite all tools being read-only. Missing readOnlyHint in tool definitions. Agents cannot distinguish safe tools from destructive ones without explicit hints.