MCP server for analyzing on-chain wallet swap transactions
Single tool with well-structured schema and comprehensive parameter documentation. Tool name follows verb_noun pattern (get_raw_transactions). Description is detailed (280+ chars) and explains the critical 'partial' field behavior. Input schema is complete with types and descriptions for all 7 parameters. However, output schema is not formally documented in the code, only inferred from the result dict construction. Error handling is present (timeout detection, partial results flag) but lacks actionable recovery guidance. No tool annotations (readOnlyHint, etc.) despite being READ_ONLY. Naming is clear and specific; parameters are well-constrained with enums and ranges implied in descriptions.
Fetch raw swap transactions for a Solana wallet address. Returns all detected swaps by default. Use optional filters to narrow results. IMPORTANT: Check the "partial" field in the response — if true, some transactions were not fetched due to timeout and results may be incomplete.
Output schema not formally documented. Tool returns a dict with 'partial', 'wallet', 'swap_count', 'partial_reason', 'filters_applied' fields, but no JSON Schema or structured type hint visible in code. LLMs cannot reliably extract fields from undocumented responses.
No tool annotations despite clear semantics. Tool is READ_ONLY (no state mutation) and idempotent (same input → same output). Missing readOnlyHint annotation that would help agents reason about safety and retry behavior.
Error handling lacks recovery guidance. Timeout errors set 'partial': true but do not suggest next steps (e.g., 'Retry with a shorter date range or smaller wallet'). Per pattern:recovery-guide, errors should tell LLMs what to do next.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | <=2025-11-25 | v2 |
Parameter 'exclude_categories' accepts array but lacks enum constraint. Description lists valid values ('transfer', 'staking', 'nft', 'other') as prose, not as a JSON Schema enum. LLMs may hallucinate invalid category names.
Parameter 'token_type_filter' uses string enum but constraint is only documented in description text, not in schema. Schema should declare enum: ['meme', 'stablecoin', null] to be machine-parseable.