PLATFORM

See it. Stop it. Before it hits production.

Detection tells you what happened. Enforcement stops it from happening. TrustScope policy enforcement runs inline with your agent traffic — simulate first, then alert, then block.

Start Free See Policy Controls

ENFORCEMENT MODES

Three modes. Your pace.

Every policy runs in one of three modes: simulate, alert, or block. Start in simulate to see what would be caught with zero production impact, then promote individual policies to alert or block as confidence grows. Each policy moves independently — there is no global switch.

Simulate

See what would be caught. Zero production impact.

“What would happen if I turned this on?”

Alert

Get notified when a policy triggers. Agent continues.

“Tell me when something fires.”

Block

Prevent dangerous actions before they execute.

“Stop it. Now.”

THREAT COVERAGE

Mapped to OWASP Agentic Top 10.

OWASP Risk	TrustScope Response
ASI-01 Goal Hijacking	Prompt injection + jailbreak AI detectors with escalation and block.
ASI-02 Tool Misuse	Command firewall, A2A depth controls, tool call policy validation.
ASI-04 Supply Chain	Schema quarantine, TOFU pinning, shadow registry tracking.
ASI-06 Context Poisoning	Context growth monitoring, guardrail policy checks.
ASI-08 Cascading Failures	Loop killer, velocity controls, cost caps, error-rate controls.
ASI-10 Rogue Agents	Agent DNA drift detection, behavioral fingerprinting, kill switch.

POLICY EXAMPLES

Policies that make sense to humans.

Budget Cap

policy: customer-bot-budget
mode: block
trigger:
  metric: session.total_cost
  operator: gte
  value: 5.00
action:
  block: true
  notify:
    channel: "#ops-alerts"
    message: "Session cost cap reached."

Command Restrictions

policy: no-destructive-sql
mode: block
trigger:
  tool: sql_query
  pattern: "DROP|TRUNCATE|DELETE\\s+FROM"
action:
  block: true
  escalate: human_review
  notify:
    channel: "#security"
    message: "Destructive SQL blocked."

PII Redaction

policy: redact-pii-in-responses
mode: alert
trigger:
  detector: pii_scanner
  entities:
    - ssn
    - credit_card
    - phone_number
action:
  redact: true
  log: true
  notify:
    channel: "#compliance"
    message: "PII detected and redacted."

HUMAN IN THE LOOP

High-risk actions wait for your approval.

When a policy triggers in block mode, TrustScope can hold the action and route it through a human approval workflow. Reviewers receive notifications via Slack, email, or webhook with the full evidence chain — the agent trace, the policy that fired, and the action that was blocked. Approve or deny with one click and the decision is logged to the audit trail.

Available at Enforce tier.

ADAPTIVE GOVERNANCE

Policies that tighten automatically.

Adaptive governance watches detection patterns over time and auto-adjusts policy constraints. If prompt injection attempts spike on a particular agent, the system can tighten that agent's enforcement mode from simulate to alert or from alert to block without manual intervention. Thresholds, cooldowns, and escalation paths are all configurable.

Available at Enforce tier.

Start with visibility. Upgrade when you need enforcement.

Monitor tier gives you detection. Protect adds blocking. Enforce adds AI-powered analysis and adaptive governance.

Start Free Compare Plans