1. Conduid
  2. Agents
  3. Hive Agent Kyc
MCP server · Agents

Hive Agent Kyc

KYA agent identity verification and trust scoring for autonomous A2A networks

Unclaimed agents
34Low

Scored 5 months ago · breakdown

About Hive Agent Kyc

Hive Agent Kyc is an MCP server in the Agents category: kYA agent identity verification and trust scoring for autonomous A2A networks.

Install

From source
git clone https://github.com/srotzin/hive-mcp-agent-kyc

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Tools (4)

agent_kyc_check_fatf_listfree

Parameters not published.

agent_kyc_check_ofac_listfree

Parameters not published.

agent_kyc_query_statusfree

Parameters not published.

agent_kyc_screen_address$0.10 USDC (Base)

Parameters not published.

Ask AI

✦

Ask AI about Hive Agent Kyc

Powered by Claude · Grounded in docs

I know everything about Hive Agent Kyc. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

  • ✓README presentThe repository ships documentation.

5 checks have not been run yet.

README


Scope disclaimer

Hive Agent KYC is a broker/observer layer. It routes screening requests to third-party providers and surfaces public sanctions list matches. It is not a regulated money-services business, does not make custody, lending, or transaction-permitting decisions, and does not issue compliance attestations on behalf of regulated entities. Final compliance determinations remain with the requesting agent and its operator.

This disclaimer is also surfaced verbatim in every tool's response payload, in /.well-known/agent.json, and in the JSON-LD description.


What this is

hive-mcp-agent-kyc is a Model Context Protocol (MCP) server that any MCP-compatible client (Claude Desktop, Cursor, Manus, agent runtimes) can call to:

  1. Route a blockchain address screening request to one of three third-party KYC/AML providers — Chainalysis, TRM Labs, or Elliptic — and return the provider's risk score and flags verbatim (no enrichment, no Hive scoring layer).
  2. Check public sanctions list matches against the OFAC SDN list and the FATF high-risk-jurisdiction lists.
  3. Read back an audit-log entry for a prior query (DID + timestamp + provider + result code only — no PII).

What this does not do:

  • It does not hold custody of any KYC documents.
  • It does not make a final allow/deny determination on a transaction.
  • It does not issue an attestation that a person/entity is "KYC'd by Hive".
  • It does not operate as a money-services business.

The purpose is to be the CLEAN-MONEY gate primitive that other Hive surfaces (Vault, Trade, Swap, Treasury) can call before settling. Compliance determinations remain with the calling operator.

  • Protocol: MCP 2024-11-05 over Streamable-HTTP / JSON-RPC 2.0
  • Transport: POST /mcp
  • Discovery: GET /.well-known/mcp.json, GET /.well-known/agent.json
  • Health: GET /health
  • Settlement: real x402 — USDC on Base L2 to wallet 0x15184bf50b3d3f52b60434f8942b7d52f2eb436e. No mock. No simulated.
  • Backend status: provider partnership keys pending — agent_kyc_screen_address returns 503 with backend_pending: true until configured. Public list tools (OFAC, FATF) return real data today.
  • Brand gold: Pantone 1245 C / #C08D23

Tools

Tool Cost Description
agent_kyc_screen_address $0.10 USDC (Base) Route screening request to Chainalysis / TRM Labs / Elliptic. Returns provider response verbatim. 503 backend_pending until keys configured.
agent_kyc_check_ofac_list free Check identifier against OFAC SDN public list. Cached 24h.
agent_kyc_check_fatf_list free Check ISO-3166 country code against FATF Call-for-Action and Increased-Monitoring lists.
agent_kyc_query_status free Read prior audit-log entry by query_id. Returns DID + timestamp + provider + result code + address hash. No PII.

agent_kyc_screen_address

{
  "name": "agent_kyc_screen_address",
  "arguments": {
    "address": "0x...",
    "chain": "base",
    "provider": "chainalysis",
    "requester_did": "did:hive:agent-vault"
  }
}

Until partnership keys are configured, returns:

{
  "ok": false,
  "status": 503,
  "backend_pending": true,
  "message": "Third-party screening backend not yet configured...",
  "x402": {
    "recipient": "0x15184bf50b3d3f52b60434f8942b7d52f2eb436e",
    "chain": "base",
    "asset": "USDC",
    "amount_usdc": 0.10
  }
}

agent_kyc_check_ofac_list

{
  "name": "agent_kyc_check_ofac_list",
  "arguments": { "identifier": "0x...", "identifier_type": "address" }
}

Returns the cache age, source URL (treasury.gov), and a deterministic match flag against the cached SDN list.

agent_kyc_check_fatf_list

{
  "name": "agent_kyc_check_fatf_list",
  "arguments": { "country_code": "IR" }
}

Returns category (call_for_action | increased_monitoring | not_listed) + FATF source URL.

agent_kyc_query_status

{
  "name": "agent_kyc_query_status",
  "arguments": { "query_id": "..." }
}

Audit log

The audit log stores only these fields per query, in memory:

  • query_id (UUID)
  • did (requesting agent's DID)
  • ts (ISO-8601 timestamp)
  • provider (chainalysis | trm | elliptic | ofac_public | fatf_public)
  • result_code (HTTP status returned)
  • address_hash (first 16 hex chars of SHA-256 of lowercase identifier)

Raw addresses, names, and PII are never stored. Operators wanting durable logs route them to their own SIEM via webhook (not a Hive responsibility).


Connect

Claude Desktop

{
  "mcpServers": {
    "hive-agent-kyc": {
      "url": "https://hive-mcp-gateway.onrender.com/agent-kyc/mcp"
    }
  }
}

Cursor / Manus / generic MCP client

Point your client at:

https://hive-mcp-gateway.onrender.com/agent-kyc/mcp

Streamable-HTTP transport, JSON-RPC 2.0, MCP 2024-11-05.

Local

git clone https://github.com/srotzin/hive-mcp-agent-kyc
cd hive-mcp-agent-kyc
npm install
node server.js
# server runs on :3000

To enable the screening provider routing locally, set any of:

export CHAINALYSIS_API_KEY=...
export TRM_API_KEY=...
export ELLIPTIC_API_KEY=...

When no key is set, agent_kyc_screen_address returns 503 backend_pending. The OFAC and FATF tools work without keys.


Settlement

Real x402 rails. The screening tool quotes:

recipient : 0x15184bf50b3d3f52b60434f8942b7d52f2eb436e
chain     : Base L2
asset     : USDC
amount    : $0.10 USDC per screen call

No mock. No simulated. No testnet pretending to be mainnet. While the screening backend is in backend_pending, no charge is applied — the x402 quote is informational so calling agents can pre-flight payment configuration.


Hive doctrine context

This server is the CLEAN-MONEY gate primitive in the Hive Civilization three-gate doctrine (NEED + YIELD + CLEAN-MONEY). Any other Hive surface that touches settlement is expected to call agent_kyc_screen_address and the public-list tools before clearing.

It is intentionally a thin broker: enrichment, scoring, and final decision remain with the third-party provider and the operator, respectively. Hive does not become a regulated MSB by routing.


License

MIT — see LICENSE.


Hive Civilization Directory

Part of the Hive Civilization — agent-native financial infrastructure.

Brand: #C08D23

README copy captured when this server was indexed; it may lag the repository. The original is authoritative.

Questions

About Hive Agent Kyc

How do I install Hive Agent Kyc?

No package has been published for this server, so install it from source: clone the repository and follow its README.

Is Hive Agent Kyc safe to use with an AI agent?

Its trust score is 34 out of 100 (low). It passes 1 of the 1 static security checks run so far. It has no ConduID identity yet, so agent calls to it are not receipted.

Is Hive Agent Kyc still maintained?

Conduid hasn't recorded a commit date for this repository yet. Check the repository directly for recent activity.