Agents/Pro

Treasury Steward

Enforces the treasury's rules, with proof.

Give it the treasury's assets and its investment policy (minimum liquidity, maximum holder concentration, maximum funding, maximum exchange inflow, mint authority, maximum exit slippage, maximum share of market cap). It tests every asset against every rule — exit slippage with a live DEX quote for the actual position size — and returns pass/fail per rule per asset, each backed by a proof id, ready to attach to a governance vote.

COMPLIANTBREACHINCOMPLETE

Built for: DAO treasuries & vaults · Risk & compliance teams

10 tADA per run on testnet · $0.25 planned on mainnet · x402, no account

Latest real run

input {"assets":[{"token":"cardano","value_usd":9000},{"token":"chainlink","value_usd":6000},{"token":"solana","value_usd":5000}],"policy":{"min_depth_usd":500000,"allow_mint_authority":false,"max_exit_slippage_pct":1,"max_position_pct_of_market_cap":0.5}}

COMPLIANT

All 3 assets ($20K) comply with the policy.

See the proof →13d ago · attested via Chainlink

How it decides

  1. 01Checks every asset against every rule: minimum depth, top-10 holders, funding, exchange inflow, mint authority, exit slippage for the whole position, and share of market cap.
  2. 02Exit slippage uses the cheaper of a live on-chain sell quote and the exchange order books.
  3. 03Native coins (ADA, SOL, ETH…) have no token contract, so no admin can mint; the mint rule passes for them.

What each verdict means

  • COMPLIANT

    Every asset passes every rule.

  • BREACH

    At least one rule fails.

  • INCOMPLETE

    No failures, but at least one rule could not be measured.

Use when: A DAO or company treasury must prove its holdings meet its investment policy.

Call it

From code, from Claude, or from any agent.

One definition, served three ways. Every run returns a verdict, a plain summary, its reasons with sources, and a proof id.

REST

POST the input, get the run. Pay per call with x402, or send a design-partner key.

curl
curl -X POST https://coingraph.ai/api/v1/agents/treasury-steward \
  -H 'content-type: application/json' \
  -d '{"assets":[{"token":"cardano","value_usd":9000},{"token":"chainlink","value_usd":6000},{"token":"solana","value_usd":5000}],"policy":{"min_depth_usd":500000,"max_exit_slippage_pct":1,"allow_mint_authority":false,"max_position_pct_of_market_cap":0.5}}'

MCP

With CoinGraph connected, Claude or Cursor calls run_treasury_steward when you ask in plain words.

tools/call
{
  "name": "run_treasury_steward",
  "arguments": {
    "assets": [
      {
        "token": "cardano",
        "value_usd": 9000
      },
      {
        "token": "chainlink",
        "value_usd": 6000
      },
      {
        "token": "solana",
        "value_usd": 5000
      }
    ],
    "policy": {
      "min_depth_usd": 500000,
      "max_exit_slippage_pct": 1,
      "allow_mint_authority": false,
      "max_position_pct_of_market_cap": 0.5
    }
  }
}

In plain words

In Claude, Cursor or any MCP client, once CoinGraph is added.

Ask
Run the Treasury Steward on our treasury rules and show me the proof.

The other nine