Connect AI assistants to ENS for name lookups, profile information, availability checks, pricing and more.
ENS MCP Server has well-structured tool definitions with mostly complete schemas and descriptions. All 7 tools are properly registered with input schemas and descriptions present. Naming follows verb_noun convention consistently (getEnsProfile, isNameAvailable, getPrice, etc.). Tool descriptions range from 84-210 characters, generally within the optimal 50-200 character range. Parameter descriptions are present and informative. However, there are several gaps: (1) no output schemas documented for any tool, (2) no tool annotations (readOnlyHint/destructiveHint) despite all tools being READ_ONLY, (3) missing error handling guidance, (4) parameter constraints could be more explicit in some cases, (5) no pagination documented for list tools despite accepting pageSize parameters. The codebase shows professional TypeScript structure with proper dependency injection via Effect, but MCP-level documentation is incomplete.
Fetch ENS profile information for a name including owner address, resolver, expiry, wrapped status, text records, address records, and content hash.
Get the complete history of events for an ENS name including domain events, registration events, and resolver events.
Get ENS names owned by an address with support for filtering by search string, expiry status, deletion status, name type, and ordering options.
Get the price for registering or renewing an ENS name for a specified duration.
Get available text and coin records for an ENS name from the subgraph.
Get subnames for an ENS name with support for filtering by search string, expiry status, and deletion status, with optional ordering and pagination.
Check if an ENS name is available for registration.
No output schemas documented for any of the 7 tools. LLMs cannot plan downstream operations or extract typed data without knowing the response structure (field names, types, required vs optional). This violates the critical 'Document the output schema' pattern.
All 7 tools are READ_ONLY but no tool annotations present. Missing readOnlyHint=true on all tools. Tool annotations (readOnlyHint, destructiveHint, idempotentHint) are current patterns in spec v2026-07-28 and enable proper agent planning (retry safety, idempotency, risk assessment).
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 55 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 48 | - | v1 |
Pagination tools (getSubnamesForName, getNamesForAddress) accept pageSize but lack documented total_count or next_cursor fields in output. Without pagination metadata, LLMs cannot iterate through large result sets or know if more results exist. Pattern requires 'Tools returning lists should accept page/offset and limit parameters and return a total count or next_cursor.'
No error handling guidance in descriptions. Tools lack recovery hints like 'If name not found, try searching with wildcards' or classification (retryable vs user-fixable). Error responses should tell the LLM what to do next.
Parameter 'duration' in getPrice uses free-form string ('1 year', '2 years', '365 days'). Should be an enum or accept structured input. Current approach invites LLM hallucination of invalid duration formats.
getEnsProfile 'textRecords' and 'coinRecords' parameters accept arbitrary arrays without documenting valid keys. Should either provide enum of common keys or clear documentation of allowed formats (e.g., 'Support standard ENS text record keys from https://...' and 'Coin symbols: ETH, BTC, SOL, or numeric SLIP44 coin types').