For AI Agents
Agents that act the instant they decide
Your agents trade, pay for services, and move money on their own. Pimlico gives them smart accounts that execute the moment their logic fires, with no native token to fund and no transactions left waiting.
Trusted by teams building AI agents
Use cases
Where agents make sense
Wherever an agent moves money or pays for something onchain, it needs an account that executes on demand, every time.
Autonomous DeFi
Agents trade, optimize yield, and rebalance positions within scoped permissions, the way ZyFAI runs its onchain strategies on Pimlico.
AI-managed treasury
Agents move and allocate funds on a schedule or a signal, against hard caps you set up front.
Agentic payments (x402)
Agents pay per API call, dataset, or model inference in stablecoins, with no API key or account to provision.
DePIN & compute
Agents buy compute, storage, and bandwidth pay-per-use, settling each request as it happens.
Agent commerce
Agents handle autonomous purchasing and recurring subscriptions on a user’s behalf.
Agent-to-agent settlement
Agents discover, hire, and pay each other for services, with settlement happening onchain.
Why agents run on Pimlico
Built for accounts that act on their own
Built for fleets of agents
Run a fleet without the upkeep
Running agents at scale usually means keeping a balance alive in every account, then watching transactions fail the moment one runs dry. That upkeep is an operational tax, and it grows with every agent you add.
Pimlico removes it. An agent’s actions are sponsored, or paid from the funds it already holds, so every agent transacts the instant its logic fires, with nothing to top up and nothing to monitor.
import { createSmartAccountClient } from "permissionless"
import { toSafeSmartAccount } from "permissionless/accounts"
import { createPimlicoClient } from "permissionless/clients/pimlico"
import { http, parseEther } from "viem"
import { base } from "viem/chains"
const pimlicoUrl = "https://api.pimlico.io/v2/base/rpc?apikey=API_KEY"
const pimlicoClient = createPimlicoClient({ transport: http(pimlicoUrl) })
const account = await toSafeSmartAccount({
client: publicClient,
owners: [owner],
version: "1.4.1",
})
// Gas is sponsored, so your user never sees a fee
const smartAccountClient = createSmartAccountClient({
account,
chain: base,
bundlerTransport: http(pimlicoUrl),
paymaster: pimlicoClient,
})
const txHash = await smartAccountClient.sendTransaction({
to: "0x...",
value: parseEther("0.1"),
})// Same client, plus one line — token approvals are injected for you
const smartAccountClient = createSmartAccountClient({
// ...same setup as before, plus:
userOperation: {
prepareUserOperation: prepareUserOperationForErc20Paymaster(pimlicoClient),
},
})
const txHash = await smartAccountClient.sendTransaction({
to: "0x...",
value: parseEther("0.1"),
paymasterContext: {
token: usdc, // any of 300+ supported ERC20 tokens
},
})Execution reliability
No human in the retry loop
Unattended accounts need execution that absorbs bursts, recovers failures, and lands every action with nobody watching.
Key custody
However your agent holds its keys, it works
The first architecture decision for an autonomous account is where its key lives. A server-side key, institutional key management through Turnkey, Fireblocks, or DFNS, a user’s passkey for user-owned agents: if it can sign, it works.
Any viem-compatible signer plugs in as the smart account owner, so your agent framework never dictates the execution layer, and switching key providers never touches your transaction infrastructure.
// Any viem-compatible signer works as an owner:
// a LocalAccount, an EIP-1193 provider, or a WalletClient.
const owner = privateKeyToAccount("0x...") // local key
// const owner = await providerToSmartAccountSigner(eip1193Provider)
const account = await toSafeSmartAccount({
client: publicClient,
owners: [owner], // Privy, Turnkey, passkeys: if it can sign, it works
version: "1.4.1",
})Spend controls
Let agents spend, within limits you set
Deploying agents on behalf of your users? Let them fund the agent, then allow it to cover its own onchain actions in USDC, USDT, or 300+ tokens, under restrictions you define. Sponsorship policies cap what you fund per agent, per action, and per campaign, scoped to specific chains and contracts. When a limit is hit, sponsorship stops on its own, so a misconfigured agent can never spend past the budget you set.
Those controls govern the cost of acting. To bound the value an agent moves on a user’s behalf, pair Pimlico with your smart account’s own guardrails: session keys and spend limits on Safe or Kernel, configurable through permissionless.js.
sp_onboarding_q3
New-user onboarding
Observability
Your agents act on their own. You still see everything.
Running an autonomous agent means answering “what did it just do?” at any moment, 3am included. Every operation is traceable in the dashboard from received to included on-chain, with millisecond timestamps and transaction hashes at each step.
Spend is itemized by policy, chain, and time, so each agent or campaign accounts for itself. Custom alerts fire before a budget or limit is hit, not after your agent stops.
Track user operation
Received
2026-06-11 20:01:13.054
Added to Mempool
2026-06-11 20:01:13.054 (+0ms)
Submitted
2026-06-11 20:01:13.341 (+287ms)
Tx Hash: 0x223b701f...Included Onchain
2026-06-11 20:01:13.590 (+249ms)
Tx Hash: 0x223b701f...Autonomous money movement runs on audited, screened, monitored infrastructure.
Frequently asked questions
What can an AI agent do with a Pimlico smart account?
Trade, pay for APIs and services, move money, settle with other agents, anything an account can do onchain. The agent acts the instant its logic fires, with no transaction it has to stop and prepare for.
Do agents need to hold a native token to transact?
No. Sponsor an agent’s actions so they run for free, or, if you deploy agents for users who fund them, let each agent cover its own fees in USDC, USDT, or any of 300+ tokens. Either way there is no native token to acquire, bridge, or keep topped up.
How do I keep an agent from overspending?
Sponsorship policies cap what you sponsor per agent, per action, and per campaign, scoped to specific chains and contracts. For a user-funded agent paying its own way, your smart account’s session keys and spend limits bound what it can spend and which contracts it can touch. When a limit is hit, the agent simply stops.
Which smart accounts and chains can agents use?
Safe, Kernel, and any account supported by permissionless.js, across 100+ chains. An agent’s smart account works exactly like any other, and switching chains is a one-line config change.