PRODUCT OVERVIEW
Evidence
Records what an agent observed, decided, and did in a tamper-evident ledger, with evidence bundles for audit review.
PRODUCT FAMILY
Compliance and Audit
BUILT FOR
Financial Services, Healthcare, Public Sector, Regulated Industries
PLATFORM
Salesforce
LICENSE
MIT
PRICING
Free
VERSION
1.0.0

THE NEED
The problem it solves
Teams need a traceable record of agent actions and the evidence behind them to support investigation and audit review.

THE PRODUCT
How it works
Evidence creates a tamper-evident decision record for Agentforce. It distinguishes observed activity, such as actions, inputs, outputs, and approvals, from the agent’s own explanation of its reasoning. A required rationale action records that explanation before consequential actions. The Salesforce-native ledger gives compliance, risk, and audit teams a clearer record to investigate without treating an agent’s explanation as proof of its internal reasoning.


CAPABILITIES
Key features
Decision_Ledger__b and Chain_Anchor__b big objects: append-only, per-agent SHA-256 hash chains over a canonicalized, key-sorted body
Daily terminal-hash anchors, the artifact a customer actually hands an auditor
Observed versus Declared separation enforced in the data model, the API and the UI, so the agent’s own account is never mistaken for a computation trace
Decision_Record__c with Rationale, Consideration and Policy_Check children driving a six-panel decision record
Sealed evidence bundles carrying their chain segment and anchor proof
Legal holds that block purge, and retention that redacts bodies while keeping the hash so the chain still verifies
Chain integrity verification passes that recompute each hash and record the result
React UI Bundle console on Multi-Framework with 10 feature routes and six least-privilege permission sets
USE CASES
Where it helps
Auditor asks what a specific agent did on a specific date and receives a sealed bundle with anchor proof
Compliance officer verifies the ledger has not been altered since it was written
Investigator reviews why humans overrode an agent, with override reasons ranked because the top one is where to fix the agent
Legal places a hold that blocks purge across a set of decision records
Risk team reviews autonomy and override rate over 30 days beside chain integrity
Retention job redacts record bodies while preserving hashes, so verification still passes