@agentvault/sdk

Minimal SDK for AgentVault policy-gated spending vaults on BOT Chain.

Three-line quickstart

const { ethers } = require(`ethers`);
const { VaultSDK } = require(`@agentvault/sdk`);

const vault = new VaultSDK(agentSigner, VAULT_ADDRESS);
await vault.spend(ethers.ZeroAddress, VENDOR, ethers.parseEther(`0.1`));

spend() dry-runs via wouldExecute first and throws a human-readable error (OverLimit, TargetDenied, …) instead of burning gas on a certain revert.

Reads

• wouldSpend(agent, token, target, amount) → { ok, reason, human }
• daySpent(agent, token), dailyLimit(agent, token)
• isTargetAllowed(target) (direct allowlist OR any bundle)
• getBundleTargets(bundleId), isOwnerMultisig()

Owner writes (use an owner signer)

setAgent, setTarget, setDailyLimit, setPerTxLimit, setCooldown, setPaused, createBundle, addTargetToBundle, removeTargetFromBundle, ownerWithdraw.

LLM-agent adapter

const { vaultSpendTool } = require(`@agentvault/sdk/langchain`);
const tool = vaultSpendTool(sdk, { agentAddress });
// LangChain: new DynamicStructuredTool(tool)
// OpenAI function calling: require(`@agentvault/sdk/langchain`).openAIFunctionSpec