Lakera Guard for AI agents
Agents read Lakera Guard at once. Changes and sends are held for a person's approval by default.
About Lakera Guard
Lakera Guard screens AI application interactions for security and safety risks and provides Enterprise APIs for managing guardrail policies and projects.
What an agent can do in Lakera Guard
Every Lakera Guard action is classed by the platform as a read, a change or a send. Reads run at once. Changes and sends are held for a person by default.
Runs at once (reads)
Looks things up in Lakera Guard without asking first. 4 actions.
- Evaluate DetectorsEvaluate content and return one confidence result per configured detector without producing an aggregate flagging decision or Guard screening log. Use for analysis and tuning, not runtime allow/block
- Get PolicyRetrieve an Enterprise Lakera policy by ID, including its detector configuration, custom guardrails, version, and assigned projects.
- Get ProjectRetrieve an Enterprise Lakera project by ID, including its assigned policy, metadata, and detect-or-flag action.
- Screen ContentScreen the latest LLM or agent interaction with Lakera Guard and return the policy decision. Use this for runtime allow/block decisions; earlier messages provide context, but only the latest interacti
Held for approval by default (changes and sends)
Held for a person's approval by default. The approval is recorded and the action is sent once, never twice, with 10 seconds to undo. 6 actions.
- Create PolicyCreate an Enterprise Lakera policy with built-in detectors and/or custom guardrails. Lakera may omit the new policy ID; if omitted, later retrieval, replacement, and deletion require external policy l
- Create ProjectCreate an Enterprise Lakera project, optionally assigning a policy, metadata, and detect-or-flag action. Omit policy_id to use the default policy. Lakera does not return the generated project ID, so l
- Delete PolicyPermanently delete an Enterprise Lakera policy by ID. The provider rejects deletion while any project is assigned to the policy.
- Delete ProjectPermanently and irreversibly delete an Enterprise Lakera project by ID. Historical requests lose their association with the project's metadata and assigned policy.
- Update PolicyReplace an Enterprise Lakera policy's complete configuration and create a new policy version. Omitted detectors or custom guardrails are removed, so retrieve the policy first when preserving existing
- Update ProjectUpdate an Enterprise Lakera project by ID. Name is required by the official request schema; retrieve the project first when preserving its current name or other optional settings.
Changes wait in Needs you
Changes and sends in Lakera Guard are held for a person's approval by default. The approval is recorded, and the action is sent once, never twice, with 10 seconds to undo.
Lakera Guard questions
Can an agent change data in Lakera Guard on its own?
Not by default. Changes and sends in Lakera Guard are held for a person's approval. The approval is recorded, and the action is sent once, never twice, with 10 seconds to undo.
What can an agent read in Lakera Guard?
This connector lists 4 read actions. Reads run at once, without waiting for approval.
How does Lakera Guard connect?
It connects with API key. Connectors are built and maintained by third-party developers; availability is confirmed during scoping.