Skip to main content

PaymentIntent

@totemsdk/agent-policy

@totemsdk/agent-policy / PaymentIntent

Interface: PaymentIntent

The action an agent wants the wallet to take. Agents produce intents; they do not execute them.

Properties​

amount?​

optional amount?: string

Amount in the token's native unit (string to preserve precision).

inference?​

optional inference?: object

Inference block — present when type === 'inference' and describes the compute the agent wants authorized. Its usage output is the metering unit budget/cap policies convert into spend.

budgetTokenId?​

optional budgetTokenId?: string

Token id the provider meters in, if inference is token-denominated.

domain​

domain: InferenceDomain

input?​

optional input?: unknown

Prompt / audio / image reference to compute over.

maxTokens?​

optional maxTokens?: number

Upper bound on output tokens, used by budget policies.

metadata?​

optional metadata?: Record<string, unknown>

model?​

optional model?: string

Requested model, when the intent targets a specific one.

op​

op: string

Provider-neutral operation name, e.g. 'completion', 'ragSearch'.

metadata?​

optional metadata?: Record<string, unknown>

Arbitrary extra context the agent wants to attach (e.g. invoice ref).

reason?​

optional reason?: string

Human-readable reason for the payment (shown to user in approval UI).

recipient?​

optional recipient?: string

Recipient Minima address (Mx… or hex).

risk?​

optional risk?: "low" | "medium" | "high"

Agent's self-assessed risk level — used by AgentPolicy routing.

tokenId?​

optional tokenId?: string

Minima tokenId, or '0x00' for native Minima.

type​

type: "payment" | "channel_update" | "settlement" | "lookup" | "receipt" | "inference"

Discriminator — what kind of operation this intent represents.