Loading ToolBench
AI Director Platform - Merchant-facing multimodal knowledge and content planning engine. Exposes core platform capabilities as MCP tools for AI agents.
OpenLucid has 14 tools with complete input schemas and descriptions. Naming follows verb_noun convention (get_, list_, create_, run_, save_). However, descriptions are inconsistent in depth, some are terse (e.g., 'List available creation apps') while others lack actionable context. The run_app tool has 21 parameters with minimal per-parameter descriptions, violating the pattern:constrained-input rule. No output schemas are documented. Error handling is absent from tool definitions. The server includes helpful MCP instructions but tool-level guidance is sparse. Average tool description length ~80 chars (below the 194-char baseline for A+ tools). Parameters lack enum constraints where applicable (e.g., app_id, action, language, content_type).
Add a knowledge item to an offer's knowledge base
Create a new merchant/workspace/brand
Create a new offer with optional knowledge inference
Get configuration for a specific app
Fetch the system prompt, user message, and output schema for knowledge inference
Retrieve merchant overview and context
run_app has 21 parameters with minimal per-parameter descriptions. LLMs cannot infer which params are required, mutually exclusive, or conditional on app_id/action values. Violates pattern:constrained-input and pattern:tool-description.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., merchant_id from list_merchants). Violates pattern:tool and pattern:response-shaper.
Descriptions are terse (avg 80 chars vs 194-char baseline). 'List available creation apps' lacks context on when to call it or what structure it reveals. Descriptions under 100 chars risk incomplete LLM understanding.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | <=2025-11-25 | v2 |
Retrieve offer context and summary information
Have OpenLucid's configured knowledge-scene model infer knowledge base items server-side
List available creation apps and their configurations
List all merchants for the authenticated user
List offers for a merchant
List strategy units for an offer
Execute a creation app (kb_qa, script_writer, topic_studio, etc.)
Save a completed piece of content (post, script, email, hook) to the user's library
Parameters lack enum constraints. 'action' in run_app, 'content_type' in save_creation, 'source_type' in add_knowledge_item should declare valid values as enums, not free-form strings. Invites hallucinated values.
No error handling guidance in tool definitions. Tools lack recovery hints (e.g., 'If merchant not found, call list_merchants first'). Violates pattern:recovery-guide.