MCP server for HiveIdentity — W3C DID resolution and agent KYC. Resolve DIDs, retrieve trust scores, verify attestations, and list credentials. OFAC/FATF screening per agent. USDC settlement on Base L2.
All 4 tools have complete input schemas with proper types and descriptions. However, tool descriptions are generic and lack actionable context for LLM selection. More critically, ALL tools return HTTP 503 with a feature-gating error message ('backend pending; submit interest at hive-mcp-connector'), they are non-functional stubs. The server explicitly violates RAILS RULE 1 by returning mock error responses instead of real data or proper error handling. Tool names follow verb_noun convention (resolve_, get_, verify_, list_) which is good. Parameter descriptions are minimal but present. No output schemas are documented. Error responses do not guide recovery or explain when/how to retry. The MPP payment rail implementation is sophisticated but orthogonal to tool quality.
Retrieve the Hive trust score (0-100) for a DID. Factors include on-chain activity, attestation count, repayment history, and peer endorsements. Backend pending (Q3 2026).
List all attestations associated with a DID — issued, received, and expired. Returns array of attestation metadata. Backend pending (Q3 2026).
Resolve a DID to its public profile: display name, verification methods, service endpoints, and Hive trust metadata. Conforms to W3C DID spec. Backend pending (Q3 2026).
Verify whether a specific attestation (by hash) was issued to a DID and is currently valid. Returns boolean. Used in KYC-lite flows and smart-contract gating. Backend pending (Q3 2026).
All tools return HTTP 503 with feature-gating error instead of real or mock data. Server is non-functional, agents cannot use these tools for any purpose.
Tool descriptions lack context for LLM selection. 'Retrieve the Hive trust score (0-100)...' does not explain WHEN to call this vs other tools, what prerequisites exist, or what the agent should do with the result.
No output schemas documented. LLMs cannot plan downstream tool calls or extract required fields (e.g., does resolve_did return verification_methods as array or object?).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
Error responses do not guide recovery. HTTP 503 with 'backend pending' tells the agent nothing actionable, should it retry? Wait? Call a different tool? No guidance.
Parameter descriptions are minimal. 'DID to retrieve trust score for' does not explain format (did:hive:0x... vs did:key:... vs other methods), validation rules, or examples.