LangGraph integration with Model Context Protocol (MCP) servers, providing tool execution nodes and multi-server MCP client adapters for agentic workflows
Two tools (add, multiply) with minimal definitions. Both have basic descriptions (10-20 chars) and identical parameter schemas (two integers). Descriptions are below the 34-char p10 baseline and lack context for LLM selection. No output schema documented. No error handling guidance. Parameter descriptions are present but generic. Tools lack verb-noun naming clarity (add/multiply are verbs but lack object context). No composition guidance or chaining IDs. Schemas are present but minimal.
Add two numbers
Multiply two numbers
Descriptions are too short (10-20 chars). Baseline p10 is 34 chars. LLMs cannot determine when to select add vs multiply without context about use cases, return format, or prerequisites.
No output schema documented. LLMs do not know what fields to expect from add/multiply results, preventing downstream tool chaining and forcing guesswork about return type.
No error handling guidance. If add/multiply fail (e.g., overflow, invalid input), LLMs receive no recovery hints. Errors should state what went wrong and suggest next steps.
Parameter descriptions are generic ('First number', 'Second number'). No constraints on range, precision, or overflow behavior. LLMs cannot validate inputs before calling.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 37 | 2026-07-28+ | v2 |
Tool names lack object context. 'add' and 'multiply' are verbs but do not clarify what is being added/multiplied. Consider 'add_numbers' or 'multiply_integers' for clarity when many tools are available.