Integrations

Connect the tool call that needs a decision.

Map the agent, its tools, and the system that performs the action. We confirm a supported route before an evaluation begins.

Integration model · coverage verified for the agreed route

Your application

Agent + tool call

Framework, model provider, and application context

KLA Policy Engine

Evaluate before execution

Allow · warn · require approval · block

Decision Desk

Required human review holds the action.

Your system

Downstream API

The caller must respect the policy and approval result.

Telemetry path

SDK and OpenTelemetry events help reconstruct activity. Enforcing a decision requires the execution checkpoint shown above.

The customer-owned-agent route is under active development. SDK instrumentation and published documentation do not establish verified enforcement for a specific integration route. Engagements use a verified route with an agreed scope.

Check agent compatibility

What we test before agreeing the route

Action boundary

Identify the tool call or API operation that must wait for a decision.

Policy outcome

Show allow, warn, require approval, and block behavior for representative requests.

Human review

Verify reviewer authority, the held action, and the result of approval or rejection.

Failure behavior

Test denied access, unavailable dependencies, retry handling, and evidence collection.

Show us the system you want to connect.

Tell us which framework, provider, and internal systems your agent uses. We will assess the route, its control coverage, and the work needed for an evaluation.

Discuss an integration
AI Agent Governance Integrations | KLA