Let agents transact.
Safely.
When an AI agent books a flight, orders supplies, or pays a vendor on behalf of a user, MondialPay provides the consent rails, spend caps, audit trail, and dispute workflow that make it safe — and auditable.
The agentic payment primitives
Building the next generation of agentic apps means giving software permission to spend money. MondialPay gives you the rails to do it safely, transparently, and at scale.
Consent envelopes
Users explicitly authorise an agent to spend up to a cap, for a purpose, within a time window — cryptographically signed.
Spend caps & scopes
Per-agent, per-purpose, per-day limits. Hit a cap and the agent stops — no surprises, no overruns.
Agent identity
Every payment is tagged with which agent initiated it, which user authorised it, and why.
Audit-ready trails
Complete transcript of agent decisions, prompts, and resulting payments — exportable for disputes.
Two-factor handshake
For high-value purchases, prompt the user for explicit confirmation before the agent commits.
Machine-to-machine
Agent-to-agent and agent-to-merchant settlement with cryptographic receipts.
Authorise once, transact a thousand times
A consent envelope is a signed grant from your end-user to your agent — scoped to a category, capped to a budget, expiring on a clock. Your agent transacts within that envelope without any further interaction, and every payment carries the proof.
See sample envelopesFour steps from prompt to receipt
Every agent payment runs through the same controlled flow — fast, programmable, auditable.
Cap, category, expiry, and 2FA threshold cryptographically signed.
MondialPay validates the envelope and runs policy checks at auth time.
Above the threshold? We ping the user for explicit confirmation.
Payment goes through, agent + user + reason logged forever.
Defence in depth, for autonomous spend
Three layers of safeguards means an agent can't quietly drain a wallet, even if the model goes off-script.
Cryptographic consent envelopes
Each envelope is signed with the user's key and validated server-side at every charge. Tampering invalidates it instantly. Users can revoke any envelope at any time from their dashboard.
Programmable spend policies
Velocity, MCC, country, and merchant whitelists enforced at auth time. Your code can also opt into a real-time webhook to make the final allow/deny decision based on its own context.
Full audit trail with reasoning
Every payment is tied to the agent's prompt, tool calls, and chain-of-thought (where you log it). Disputes become trivial because you can show exactly what the agent saw and decided.