Loading ToolBench
Provides real-time SEO intelligence for AI coding tools. Integrates with Google Search Console, performs site crawls, generates structured data, and provides content optimization recommendations.
Server has 19 tools with visible schemas and descriptions, but quality is inconsistent. Naming follows verb_noun convention well (get_*, create_*, etc.), but many tools have generic or incomplete descriptions that lack context about when to use them, error handling, or what happens on retry. Parameter descriptions vary widely, some are clear, others lack format guidance or constraints. Critical issue: descriptions like 'Get SEO issues for entire site' are only ~40 chars, below the 50-char baseline for LLM comprehension. No output schemas are documented; LLMs cannot plan downstream tool calls. Error handling is mentioned in the rubric but not visible in the source, no recovery guidance in tool descriptions. Composition is weak: feature specification tools (create_feature_spec, get_feature_spec, update_feature_spec, link_commit, get_commit_message) and keyword cluster tools (create_keyword_cluster, get_keyword_clusters, create_content_spec) form tangled dependency chains not described in tool docs. The server also makes heavy use of optional domain/project_id parameters with fallback to environment variables, which is undocumented and creates silent dependency risks.
Trigger a fresh site crawl and analysis. Use this after deploying changes to refresh SEO data. Crawls the entire site, checks all URLs, detects issues, and updates the database with current SEO status.
Create a content specification from a keyword cluster.
Create and save a feature specification to Rampify. IMPORTANT: Before calling this tool, YOU (Claude) must generate the complete structured spec from the user's description and your codebase context. Do not pass raw natural language — populate all fields: Infer affected_files from open files and the codebase structure; Infer tech_stack from package.json and imports; Generate 3-5 acceptance criteria covering happy path, edge cases, and error handling; Break implementation into 3-8 concrete tasks with file references; Write ai_context_summary to help future AI agents understand the approach; Set next_action to the single most important first step
Create a keyword cluster for content planning.
Generate optimized meta tags (title, description, OG tags) for a page. Analyzes page content and provides recommendations for SEO-optimized meta tags based on actual content, headings, and topics.
Output schemas not documented. Tools return JSON but LLMs cannot see what fields to expect, breaking downstream composition and forcing agents to guess field names.
Descriptions lack actionable detail. Most are 40 - 60 chars and do not explain WHEN to use the tool vs similar ones, what happens on retry, or how to handle errors. E.g. 'Get SEO issues for entire site with health score' does not explain whether this is for initial audit or ongoing monitoring.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 58 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 36 | - | v1 |
Auto-generate structured data (schema.org JSON-LD) for any page. Detects page type (Article, Product, FAQ, etc.) and generates appropriate schema with validation. Returns ready-to-use code snippets.
Generate an optimized git commit message based on a feature spec.
Get a previously created feature specification by ID.
Get Google Search Console performance insights with AI-powered content recommendations. Returns top performing pages, query opportunities (improve CTR, rankings, keyword gaps), and actionable recommendations for what content to write next.
Get SEO issues for entire site with health score. Returns health score (0-100), categorized issues by severity, and prioritized recommendations for fixes.
List keyword clusters for a project.
Get comprehensive SEO data and insights for a specific page. Returns performance metrics from Google Search Console, detected issues, optimization opportunities, and actionable recommendations.
Get security and compliance context for the site.
Link a git commit to a feature specification.
List all feature specifications for a project.
Look up keyword data and search volume information.
Get AI-powered content optimization recommendations for a page.
Get AI-powered keyword suggestions related to a topic.
Update an existing feature specification.
Optional domain/project_id parameters silently fall back to env vars without documentation. LLMs cannot know whether they must pass these or if the server will use defaults. Silent dependency creates unpredictable behavior.
Feature spec and keyword cluster tools form tangled multi-step workflows (create_feature_spec → link_commit → get_commit_message; create_keyword_cluster → create_content_spec) but dependencies are not documented in descriptions. Agents cannot see that create_content_spec requires cluster_id from create_keyword_cluster.
No documented error handling or recovery guidance. Tools like crawl_site and create_feature_spec can fail, but descriptions do not say what to do on error or whether retries are safe.
Parameter descriptions lack format/constraint guidance. E.g. 'period' enum in get_gsc_insights is described as 'Time period for analysis' but does not explain why only 7d/28d/90d are valid or what happens if you request 14d.
create_feature_spec description warns agent to generate structured spec before calling, but no tool exists to validate or auto-generate specs. Risk: agents pass raw user input instead of structured spec.
Tools accept both domain and project_id but relationship is unclear. Can you call create_feature_spec with project_id only? If domain is missing, does it error or use env var? Documentation needed.