Virtuals ACP (acp-cli)

84Agent Risk Score84% of the points we could check · partial, 6 of 7 inputsHow much an AI agent could lose here if it is wrong, hacked or prompt-injected. Higher is safer.

Last checked 1 official sources linkedHow we scoreIncident tracker

Category
Agent tokens
Pricing
not stated (providers set job prices)
Best for
Agent tokens · USDC agent-to-agent job payments held in escrow, agent wallets (EVM and Solana), virtual payment cards
Verified
2026-10-26

Visit Virtuals ACP (acp-cli)

Official site (source): https://whitepaper.virtuals.io/about-virtuals/commerce-layer

Agent Risk Score (v2)

Agent risk 84% (partial, 6 of 7 inputs) If an agent using this tool is wrong, hacked or prompt-injected, how much money can it lose? Higher is safer. Each input links to the vendor's own docs; affiliate links never affect it. How it's calculated · Incidents

Missing: Transaction preview or simulation. Missing inputs lower the maximum instead of counting as zero.

InputFrom official docsPoints
Who holds the keysAgent wallets are Privy-managed; signer keys are kept in the OS keychain after browser approval, and Node.js never touches raw key material at rest source10 / 20
Spending limits enforced outside the agentWallet policies are allowlists of contract or wallet addresses, enforced server-side on every transaction the signer makes; editing them needs the dashboard source20 / 20
Human approvalOptional deny-all signer policy: manual approval for every transaction (the default restricted policy authorizes all ACP transactions without asking) source15 / 15
Sending funds to new addressesUnder the deny-all policy nothing leaves the wallet without manual approval; otherwise wallet policies limit destinations to an address allowlist source15 / 15
Transaction preview or simulationMissing: no official source found yetmissing / 10
Open source and auditsacp-cli is open source (ISC); no audit report found for ACP source6 / 10
Incident record (last 12 months)No incident naming this tool in our incident tracker for the last 12 months, and every scored input links an official source source10 / 10
Total84% of the points we could check76 / 90

The Safety score (v1) isn't shown in this section: it is built for tools that place orders. Why

Full facts

CategoryAgent tokens with a product
AssetsUSDC agent-to-agent job payments held in escrow, agent wallets (EVM and Solana), virtual payment cards
Live vs paperlive on Base mainnet (ACP v2.0, April 2026); Base Sepolia and BSC testnet also supported
Order approval / safetyagent wallets are Privy-managed and signer keys sit in the OS keychain after browser approval; signer policy is restricted by default (all ACP transactions), deny-all needs manual approval for every transaction, unrestricted needs none; wallet policies are address allowlists enforced server-side on every transaction and changed only in the dashboard; cards carry a spend limit; job payments stay in escrow until the client completes the job (released to the provider) or rejects it (refunded)
Pricing / free tiernot stated (providers set job prices)
Rate limitsnot stated on the pages we checked
Official / communityofficial
Last checked2026-10-26
Sourcegithub.com
Repositoryhttps://github.com/Virtual-Protocol/acp-cli

Change log

Safety check: Is Virtuals ACP (acp-cli) safe to connect to an AI agent?

Recent changes

All changes across tools

Related: Agent tokens

Are you the maker? Claim this listing (adds a separate "Provided by vendor" box; it never changes our sourced facts), Add a "Listed on TradingAgentStack" badge or send a correction.

More in Trade with a bot

Other paths: Build an AI agent · Markets & infrastructure · Learn · Search all tools

Get the weekly agent risk update

New Agent Risk Scores, incidents and safety changes for the wallets, exchanges and brokers AI agents use. One email a week, unsubscribe anytime.

No spam. Unsubscribe anytime. Privacy