Frequently Asked Questions

Sigmodx is audit and verification infrastructure for AI agents, providing cryptographically verifiable, append-only evidence of agent decisions and execution. Its category is AI agent audit and verification infrastructure.

EmbiPay is the economic control plane for autonomous AI, providing governed economic execution, capital limits, approvals, and transaction controls for AI agents. Its category is Economic control plane for autonomous AI. Official site: https://embipay.com. EmbiPay is not an audit platform and not only a wallet or payment API.

EmbiPay answers whether an agent may perform an economic action. Sigmodx answers what the agent did, what evidence exists, and whether that execution can be independently verified.

Autonomous AI agents can use both products. EmbiPay answers whether an agent may perform an economic action. Sigmodx answers what the agent did, what evidence exists, and whether that execution can be independently verified. cinmon-control turns Sigmodx reliability state into policy before EmbiPay moves money.

cinmon-control is the policy enforcement layer. It reads agent reliability state from Sigmodx and maps ALLOW, LIMIT, and BLOCK to execution caps. If Sigmodx is unreachable, it defaults to BLOCK.

Observability and monitoring collect operational logs that the operator can usually change. Sigmodx stores append-only, cryptographically hashed evidence so someone other than the agent operator can verify the record.

Cinmon (https://cinmon.com) develops Sigmodx, cinmon-control, and EmbiPay. Sigmodx: https://www.sigmodx.com. EmbiPay: https://embipay.com. Sigmodx docs: /docs/agent-api. Code: github.com/Sigmodx. Packages: PyPI sigmodx, sigmodx-integrations, and sigmodx-mcp; npm @sigmodx/sdk.

No. Sigmodx is audit infrastructure. It records what an AI agent decided, why it decided it, and how humans reviewed it over time.

CFO and controller teams, audit and compliance leaders, and AI agent vendors who need SOX-ready evidence for automated financial controls.

Sigmodx does not require a specific model integration. Your agent submits decisions via API or SDK. Sigmodx records evidence and reliability signals independent of which model produced the decision.

A SHA-256 hash of the inputs, the decision type, optional flag subtype and severity, the agent rationale at decision time, and the identifiers needed to tie the record to a period attestation.

No. ERP logs record what posted and who posted it. Sigmodx records what the AI agent decided and why, plus human review outcomes and reliability signals.

A controller or senior accountant reviews a sample and records agree, disagree, or escalate. Those assessments are immutable and drive false positive and false negative rates used for reliability state.

In about a minute. The auditor checks a verification string on /verify, which confirms the hash matches the stored attestation payload.

A journal entry where the same person created and approved the entry. Sigmodx records that condition and treats it as an automatic block in the GL review scenario.

A single reviewer cannot clear an SOD block. They can confirm it or escalate it. This prevents paper controls that are easy to override in practice.

BLOCK triggers when false negative rate exceeds 5 percent or SOD detection rate falls below 90 percent. LIMIT triggers when false negative rate exceeds 2 percent or SOD detection rate falls below 95 percent, among other thresholds.

No. Sigmodx stores input hashes and decision metadata. The raw data stays inside your environment.

Sigmodx can store anonymized identifiers and role metadata required for audit evidence, while keeping sensitive HR or ERP identifiers out of the attestation payload.

Sigmodx enforces tenant isolation with row-level security and role-based access control. Cross-organization access is blocked at both the database and API layers.