Server has 5 tools with consistent naming (verb_noun pattern: get-*) and descriptions present for all. Input schemas are visible and properly typed with Zod. However, descriptions are generic (avg ~100 chars, below 194-char baseline), parameter descriptions lack actionable detail (e.g., 'Blockchain address to query' doesn't explain format or validation), and output schemas are not documented. Error handling exists but is minimal. No tool annotations (readOnlyHint, etc.) despite all tools being read-only. Composition is sound, each tool has one responsibility.

• Document output schemas for all tools. Define the structure of the text response or refactor to return structured JSON (e.g., {transaction: {id, type, status, date, inputs, outputs, fees}} instead of formatted text). This enables LLMs to extract and chain data.
• Add explicit constraints to parameter descriptions. E.g., 'address: Blockchain address (40-42 hex chars for Ethereum, 26-35 alphanumeric for Bitcoin). Example: 0x1234... or 1A1z7agoat...' Specify valid formats, lengths, and examples.
• Enhance tool descriptions to 80-150 chars explaining WHEN to use each tool. E.g., 'get-transaction: Retrieve swap details, fees, and status for a single transaction. Use when you have a transaction hash. Returns input/output assets and liquidity fees.'
• Add readOnlyHint: true to all 5 tools in the schema. This signals to agents that these calls are idempotent and safe to retry without side effects.
• Improve error messages with recovery guidance. E.g., 'Transaction not found. Verify the txid is a valid 64-char hex string. If the transaction is recent, it may still be pending confirmation, try again in 30 seconds.' Classify errors as retryable vs user-fixable.