Capabilities

Each capability is a decision type with its own input and its own result, under the same envelope. Call them through POST /api/v1/decide with type, or through their own endpoint; both run the same engine.

Capability	Question it answers	Decision from	Own endpoint
Payment	Let this payment through?	fraud probability vs your cut-offs	`POST /assess`
Agent intent	Is this agent action within its mandate?	seven decision gates + two system steps	`POST /orchestration/intent`
Agent (KYA)	Trust this AI agent now?	KYA Score and risk flags	`POST /kya/verify`
Screen	Is this party listed?	match strength, DOB and country agreement	`POST /screen`
Document	Is this ID consistent and unedited?	MRZ check digits + alteration score	`POST /documents/check`
Statement	What AML and fraud risk is in this history?	risk band from the statement matrix	`POST /statements/analyze`
Account takeover	Was the account taken over first?	takeover likelihood vs holder evidence	`POST /ato/review`
Credit	Can this customer carry the obligation?	affordability band	`POST /credit/analyze`
Merchant	Onboard or keep this merchant?	merchant risk band	`POST /merchants/review`
KYC	Is this applicant verified?	reasons: decline > review > verified	`POST /kyc/verify`
Ask	Questions about a result	trained intent router	`POST /ask`

Composing capabilities

Capabilities are atomic; compose them in your code. A typical onboarding runs kyc (which already screens the name and checks the document), then statement or credit on the first statements; a payments flow runs payment on every authorization and ato on every fraud dispute.

kyc = client.kyc.verify(applicant, email_challenge_id=e, phone_challenge_id=p)
if kyc.status == "VERIFIED":
    risk = client.decide("statement", {"files": first_statements, "subject_ref": applicant_id}, subject_consent=True)

Availability

Each capability is enabled per account. A capability that is not enabled for your key answers 404, through /decide too.