Run Scan
Running security probes in Akto involves assessing agent components to evaluate their security posture. You can run probes instantly or schedule them for later, making it useful for planning probing sessions or implementing regular security validation.
Running Probes on Agent Components
Navigate to Agentic Discovery > Collections
Select the collection containing the agent components you want to probe
Choose specific agent components or select all components in the collection
Click Run Scan
Select probe categories to run (Prompt Injection, MCP Security, LLM Security, etc.)
The Scan panel displays the estimated token usage for every probe. Token estimation helps enterprise teams evaluate expected LLM consumption before starting the scan.
Configure probe parameters if needed.
Scan Execution Parameters
Run Type
Defines how the scan runs. You can run the scan once or configure scheduled execution.
Select Date
Defines the date when the scan execution starts.
Select Time
Defines the time when the scan execution begins.
Scan Run Time
Defines the maximum duration allowed for the scan execution. The scan stops automatically when the configured duration expires.
Select Scan Role
Associates a predefined scan role that defines permissions and execution behavior for the scan.
Max Concurrent Requests
Controls the number of parallel probe requests executed during the scan. Higher concurrency increases scan speed but also increases request load.
Max Agent Tokens
Defines the maximum number of tokens that the agent can consume during the scan. The No limit option allows unrestricted token usage.
Slack Alert Integration
Enables Slack notifications after scan completion. The Slack integration sends alerts for scan results.
Microsoft Teams Alert Integration
Enables Microsoft Teams notifications after scan completion. The Microsoft Teams integration sends alerts for scan results.
Click Run once now to start the security scan.
Probe Categories
When running probes, you can select from specialised categories:
Prompt Injection: Probes for direct and indirect prompt injection vulnerabilities
LLM Security: Validates model-specific security controls
MCP Security: Probes Model Context Protocol implementations
Sensitive Information Disclosure: Detects credential and PII exposure
Excessive Agency: Validates tool permission boundaries
Supply Chain: Identifies dependency vulnerabilities
and more...
Modify Probe Parameters
You can add, modify, or remove headers and parameters directly from the probe configuration without editing YAML files. This allows you to customize probes for specific agent behaviors or authentication requirements.
Last updated