Runtime Proof
From agent action to auditable decision
Aegisora sits between your AI agents and the systems they touch. Each action is evaluated before it runs and leaves verifiable evidence behind.
- ALLOW
crm.read_ticket
support-agent · Within scope
- ESCALATE
payments.transfer $48,000
finance-agent · Above approval limit
- BLOCK
db.drop_table users
code-agent · Destructive action denied
How it works
Four steps from request to evidence.
Agent action
An agent requests a tool call or data access through the Aegisora API.
Context and risk
Aegisora resolves the workspace from your API key and reads the agent, action, resource and the risk value you send.
Policy decision
Policies return ALLOW, BLOCK or ESCALATE, with sensitive data masked.
Evidence
Every decision is written to a hash-chained evidence record for audit.
Response
Illustrative decision
Shape of a real response, values abbreviated. The request format is in the quickstart.
{
"effect": "BLOCK",
"reason": "Policy: no external transfer of customer PII",
"policy_id": "<uuid>",
"evidence_hash": "<hash>"
}Run it against your own action.
Create a test workspace, send one request and read the verdict.