Name: Veto
Author: Plaw Labs

Building AI agents for multiple tenants requires careful isolation. One tenant's agent should never access another tenant's data.

Tenant Isolation

tenant_isolation.pypython

from veto import Veto, Policy

veto = Veto(api_key="veto_live_xxx")

async def handle_request(request, tenant_id: str):
    # Set tenant context for this request
    with veto.tenant_scope(tenant_id):
        decision = await veto.validate({
            "tool": "query_database",
            "arguments": request.args,
        })

# Per-tenant policies
veto.register_tenant_policy(
    tenant_id="tenant_123",
    policy=Policy(
        tool="query_database",
        rules=[
            Policy.allow_if(table_prefix="tenant_123_"),
            Policy.deny_if(table_prefix="tenant_"),
        ]
    )
)

Best Practices

Always validate tenant context before tool execution
Use tenant-scoped policies
Log tenant ID with every decision
Rate limit per tenant, not globally