Agentic Credit Broking ProtocolPrecise terminology explained in plain language. For formal definitions, see the whitepaper on GitHub.The Agentic Credit Broking Protocol uses precise terminology to define roles, actions, and responsibilities. This page explains the key terms in plain language. For formal definitions, see the whitepaper on GitHub.
Read the whitepaper on GitHub →An open standard developed by ClearScore that enables AI assistants to participate in regulated credit broking journeys — while brokers retain full regulatory control and lenders receive a structured, auditable application.
The structured, replayable record of every action in an AI-mediated credit journey. Includes User Agent identity, structured financial data provided, broker actions and resolutions, and conversation transcripts where available. The broker maintains this record and is responsible for it.
The process by which a User Agent's protocol compliance is tested against standardised scenarios and recorded in a public participant registry. Certification means a User Agent has demonstrated it can handle the protocol's operations correctly — including presenting regulated content verbatim and capturing consent explicitly.
A typed instruction the broker issues to the User Agent — such as a disclosure, consent request, declaration, or instruction — that requires an explicit response. Broker actions with a "regulated" flag must be presented to the user verbatim, without paraphrase or omission.
An explicit permission the user must grant before the broker can take a specific action — such as running a credit search or sharing data with a lender. Consent cannot be inferred from conversational agreement. The user must respond explicitly to a typed consent action.
The regulated firm that assesses the user's financial circumstances, identifies suitable credit products, and is responsible for compliance throughout the journey. In ACBP, the broker controls the compliance gates and maintains the audit trail — regardless of which AI assistant mediates the interaction.
A regulated statement the broker must communicate to the user verbatim. The AI presents it exactly as supplied — it must not summarise, rephrase, or omit the content. The user must explicitly acknowledge it before the journey continues.
A compliance-critical step in the journey that the AI cannot bypass, infer from conversation, or skip. The broker issues a blocking gate as a typed action. The case does not advance until the gate is explicitly resolved and logged. Examples include Disclosure, Consent, Declaration, and Instruction.
The credit provider that makes the final credit decision. The lender's relationship is with the broker, not with the AI assistant. The protocol does not change how brokers and lenders interact — it changes how users reach brokers.
The framework by which brokers calibrate how much they rely on a given User Agent. Unknown platforms start as introducer-only. Efficiency — including the ability to handle regulated gates directly — is earned through demonstrated compliance over time.
The AI assistant or application through which the user interacts during the journey. The User Agent mediates the conversation — gathering information, presenting offers and disclosures, executing the broker's instructions. It is not a regulated entity and does not carry regulatory responsibility.
These definitions are intended as plain-language summaries. For formal definitions and full technical detail, read the whitepaper. Read the whitepaper on GitHub →