RWA Pipe MCP — RWA Data for AI Agents

The first MCP server purpose-built for AI agents that need to query, route, or hold tokenized real-world assets. ~$30B in distributed asset value across treasuries, private credit, real estate, gold, and equities — exposed as 82 tools any Claude / Cursor / Continue / custom agent can call.

Designed for the AI-treasury-management story: agents holding idle USDC need yield destinations they can evaluate and allocate to under explicit compliance constraints. RWA Pipe gives them the data layer to do that.

Why AI agents specifically?

Nobody else does this for RWA. Coinbase / BitGo / CoinGecko have crypto MCPs. The Graph is generic. rwa.xyz sells dashboards to humans for $500/seat. We are the only canonical RWA MCP — and we shipped it before the category exists.

Quick Start

npx rwapipe-mcp

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "rwapipe": {
      "command": "npx",
      "args": ["rwapipe-mcp"],
      "env": {
        "RWAPIPE_API_KEY": "rwa_free_your_key_here"
      }
    }
  }
}

Get a free key at https://rwapipe.com/register. Free tier = 100 req/min, no credit card.

Cursor / Continue

Same MCP-server config block under your editor's MCP settings. The server speaks both stdio and SSE.

Flagship tools for agent use cases

get_rwa_tokens — constraint-driven token discovery

The whole point: agents specify what they need in plain JSON and get a list of matching tokens.

get_rwa_tokens({
  category: "us-treasury",
  kycRequired: false,         // permissionless only
  audited: true,
  redeemable: true,           // must have on-chain exit
  minTvl: 50_000_000,         // institutional-grade
  hasApy: true,
  sortBy: "tvl",
  limit: 10
})

14 constraint filters map directly to the questions a treasury allocator agent has to answer:

Plus the basics: chain, category, issuer, search, hasApy, excludeStablecoins, distributionType, platform, eligibleInvestors.

get_token_risk_signals — single-call risk dashboard

Returns the structured signals an agent needs to decide "is it safe to hold this right now":

{
  "token": {"chain":"ethereum","symbol":"BUIDL","issuer":"BlackRock"},
  "signals": {
    "tvl": {"current_usd": 147652591, "change_24h_pct": 0.01, "anomaly": "stable"},
    "mint_burn_24h": {"net_qty": 0, "anomaly": "quiet"},
    "whale_activity_24h": {"large_transfers_count": 0, "anomaly": "quiet"},
    "attestation": {"url": null, "status": "missing"},
    "compliance": {
      "kyc_required": true, "accredited_only": true,
      "jurisdiction": "US", "transferable": false,
      "audited": true, "custodian_type": "qualified"
    }
  },
  "flags": ["no_attestation_url", "kyc_required", "accredited_only"]
}

Anomaly classifications (stable/rising/falling/volatile for TVL; quiet/expansion/contraction for mints; quiet/active/outflow_spike for whales) replace the agent having to threshold raw numbers itself. The flat flags[] array is designed to be matchable against an agent's allow/deny rules.

get_whale_alerts — real-time whale-flow stream

Real-time mint/burn/transfer events >$50K USD with 1h delay on free tier. Agent subscribes to its own filter (chain, category, symbol).

Polling pattern for "subscribe to changes"

MCP servers don't push — agents poll. Recommended cadence for a treasury-monitoring agent:

Compare against the agent's last-seen state, alert on diffs. Free tier (100 req/min) is plenty for this pattern across dozens of tokens.

Reference use case: "treasury allocator agent"

Prompt to Claude with this MCP loaded:

"I have $10M idle USDC on Base. Find the best yield-bearing tokenized treasury exposure with these constraints: no KYC, transferable on-chain, contract audited, minimum $50M TVL, redeemable within T+2. Rank by APY and current risk signals. Show your reasoning."

What the agent does:

1. Calls get_rwa_tokens({chain: "base", category: "us-treasury", kycRequired: false, transferable: true, audited: true, minTvl: 50_000_000, hasApy: true, sortBy: "apy"})

2. For each candidate, calls get_token_risk_signals(chain, address) to check anomalies

3. Filters out tokens with flags matching no_attestation_url or whale_outflow_spike_24h for added safety

4. Returns ranked list with explicit risk reasoning

This whole loop is one user prompt and three MCP calls — no website scraping, no PDF reading, no broker. That's the AI-RWA opportunity.

All 82 tools

AI-agent decision-making (new, flagship)

Tokens

Market

TVL

Issuers, Categories, Chains

Mint/Burn

Risk & Whales

Yields, Webhooks, Contracts

Supported blockchains (22 active)

EVM

ethereum, arbitrum, avalanche, base, bsc, celo, gnosis, linea, mantle, optimism, polygon, zksync, fraxtal, ink, plume, scroll, fantom, sei

Non-EVM

solana, stellar, hedera, sui, ton, xrpl, near

(algorand / tron deprecated 2026-04-30 due to indexer freshness — chain handlers retained for future re-add.)

Asset categories

Environment variables

SSE mode (hosted)

For multi-tenant agent platforms, run as SSE server:

MCP_MODE=sse MCP_PORT=3001 npm start

Endpoints: GET /sse (connection), POST /message (RPC), GET /health, GET / (info).

A hosted SSE server is also available at https://rwapipe.com/mcp for low-volume experiments.

Roadmap (pre-1.0)

• ✅ 82 tools, 22 chains, 1004 tokens (2026-04-30)

• ✅ AI-agent constraint filters + risk-signals composite

• 🚧 LLM-parsed attestation summaries (get_diligence_summary)

• 🚧 Token taxonomy: regulatory_class, programmability, risk_class

• 🚧 Agent-subscription stream (subscribe_anomaly)

• 🚧 x402 metering — pay-per-call in USDC

Links

• Website & API: https://rwapipe.com

• API docs: https://rwapipe.com/docs/api

• MCP docs: https://rwapipe.com/docs/mcp

• GitHub: https://github.com/rwapipe/mcp

License

MIT