Skip to main content

Adding Governance

This tutorial covers adding governance policies to control and audit agent behavior.

What is Governance?​

Governance in Waxell includes:

  • Rate Limiting: Control execution frequency
  • Approval Workflows: Require human approval for sensitive actions
  • Audit Logging: Track all agent actions
  • Policy Enforcement: Apply rules to agent behavior

Rate Limiting​

Limit how often an agent can execute:

@agent(
    name="email-sender",
    rate_limit={
        "requests_per_minute": 10,
        "tokens_per_minute": 50000
    }
)
class EmailSender:
    pass

Approval Requirements​

Require human approval for sensitive operations:

@tool(requires_approval=True)
def delete_account(self, ctx, user_id: str):
    """Delete user account - requires human approval."""
    pass

@tool(
    requires_approval=True,
    approval_policy="two_person_rule"
)
def transfer_funds(self, ctx, amount: float, destination: str):
    """Transfer funds - requires two approvers."""
    pass

Audit Logging​

All agent actions are automatically logged:

# View audit logs via CLI
# wax agent trace <execution_id>

# Or query programmatically
from waxell_infra.telemetry import get_execution_trace

trace = get_execution_trace(execution_id)
for event in trace.events:
    print(f"{event.timestamp}: {event.action}")

Policy Configuration​

Configure policies in the control plane:

# policies/email-sender.yaml
agent: email-sender
rules:
  - name: rate-limit
    type: rate_limit
    config:
      requests_per_minute: 10

  - name: content-filter
    type: content_filter
    config:
      blocked_patterns: ["password", "ssn", "credit card"]

  - name: require-approval
    type: approval
    config:
      actions: ["delete_account"]
      approvers: ["admin"]

Next Steps​