As your crew of agents grows, so does its reach. Kynara governs which agent and task can call which tool — with approvals for high-risk actions and a full audit trail.
Multi-agent crews multiply the surface area: more agents, more tools, more autonomous handoffs. Without per-action authorization, any agent in the crew can use any tool it's wired to — and a compromised or misdirected agent can do real damage.
from kynara_sdk import permission_required
# Guard the tools your crew's agents use
@permission_required("email.send", resource_arg="recipient")
def send_email(recipient: str, subject: str, body: str):
return mailer.send(recipient, subject, body)
# Kynara enforces non-escalation across agent -> user delegation,
# so a crew member can never exceed the user who dispatched it.Denied calls are blocked before they run; high-risk actions can require human approval. Every decision is recorded in a tamper-evident audit log.
Every consequential call is evaluated against RBAC + ABAC policies and runtime context before it executes.
An agent can never exceed the permissions of the user it acts on behalf of.
Route destructive or external-facing actions to a human for approval, with full context.
Every decision is appended to a SHA-256 hash-chained log for forensics and compliance.
Go deeper: read the guide · docs · compare Kynara.
New guides on AI agent governance, MCP security, and compliance — no spam.