mizara 0.1.0

Mizara — programmable authorization layer for AI actions

mizara

Authorization layer for AI agents. Evaluates whether an agent action should proceed — before it executes — against a policy you define. Returns a signed decision receipt for every call.

pip install mizara

from mizara import create_mizara_client

mizara = create_mizara_client(policy_path="./policy.json")

result = mizara.authorize(
    actor={"id": "agent_support_v4", "type": "autonomous_agent", "framework": "langgraph"},
    action={"name": "execute_refund", "risk_profile": "high_irreversible"},
    resource={"type": "monetary_transaction", "id": "tx_99210", "attributes": {"amount": 75, "currency": "USD"}},
    context={"client_id": "acme_corp", "target_jurisdiction": "EU"},
)

# result.status  -> "ALLOW" | "DENY" | "REDACT" | "RE_ROUTE"
# result.cryptographic_receipt -> CryptographicReceipt(id, hash, signature)

Policies are plain JSON — no Rego, no Cedar syntax. See the TypeScript SDK for policy format docs.

License

Apache-2.0