Approvals
Risky steps wait for a person; you choose which, per task.
Security
Autonomous by design. Governed by you.
Every agent acts with the asker's own access, risky steps wait for a person, and every step is written to an audit log you can verify.
Controls
How every agent is governed
Policies
Cedar policies decide which person or agent may do what.
Guardrails
Every model input and output checked for prompt injection, PII and denied topics.
Budgets and kill switches
Spend capped per agent, team and tenant; stop one agent or the whole fleet.
Tamper-evident audit
Every call, cost and decision in a hash-chained log you can verify and export.
The asker's own access
Agents act with the permissions of the person who asked, never more.
Identity
OIDC single sign-on (Entra ID, Okta, Google) and SCIM 2.0 provisioning.
Self-hosting
Under an enterprise licence, in your own cloud or on-premises, with the model you choose.
Documents
Read the details
Privacy policy
What we collect, why and for how long: liyagent.com/privacy.
Data Processing Addendum
GDPR Art. 28 processor terms: liyagent.com/dpa.
Sub-processors
Who processes data for the hosted service: liyagent.com/subprocessors.
Data handling
What leaves your network and what never does: liyagent.com/docs/govern/data-handling.