MCP server for HiveInsurance — parametric insurance for autonomous agents. Uptime SLA, treasury slippage, and oracle-failure coverage. Broker layer operational; underwriter layer in development. USDC settlement on Base L2.
All four tools have complete input schemas with proper JSON Schema structure and type definitions. However, descriptions are severely compromised by feature-gating language ('Backend pending Q3 2026') that obscures actual tool purpose. Tool names lack action verbs (quote_policy, bind_policy, claim_status, parametric_trigger_check are noun-heavy or passive). Parameters have descriptions but lack enums for constrained values (risk_type should enumerate valid categories). No output schemas documented. No error handling guidance beyond generic 503 feature-gate response. Tools are read-only or write operations but lack explicit destructiveHint/readOnlyHint annotations. The server is essentially a scaffold with no functional backend, making it unsuitable for production use.
Bind a previously quoted policy. Settlement via x402 / USDC on Base, Ethereum, or Solana. Returns policy_id on success. Backend pending (Q3 2026).
Check the status of a parametric insurance claim associated with a policy. Returns current state: pending / triggered / paid / disputed. Backend pending (Q3 2026).
Check whether a parametric oracle event hash satisfies the trigger condition for automatic claim payout. Returns boolean. Backend pending (Q3 2026).
Quote a parametric insurance policy. Returns a premium estimate for the given risk type, coverage amount (USD), and term length. Backend pending (Q3 2026).
All tool descriptions are dominated by feature-gating language ('Backend pending Q3 2026') that obscures actual tool semantics. LLMs cannot determine when to call these tools or what they actually do.
Tool names lack action verbs. 'quote_policy' and 'bind_policy' are noun-heavy; 'claim_status' is passive. Should be 'get_quote', 'create_policy', 'get_claim_status', 'check_trigger_condition' to signal intent.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields (e.g., quote_id from quote_policy, policy_id from bind_policy).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2026-07-28+ | v2 |
risk_type parameter accepts free-form strings with only an example ('flight_delay, crop_yield, protocol_hack, weather'). Should be an enum to prevent hallucinated values.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). bind_policy is destructive (creates policy); claim_status and parametric_trigger_check are read-only. LLMs need this metadata.