Skip to main content
Launching March 1, 2026
Home Features Pricing Governance Security Contact
Join Waitlist

Governance

Policy-driven AI control

Define what agents can do, which models they use, how much they spend, and where data flows. Every run governed, every action audited.

How policy evaluation works

Every agent request passes through the policy engine before execution. Policies cascade from organization to project to agent.

Agent Request
User sends message or triggers action
Policy Evaluation Engine
Org Policies
Global rules
Project Policies
Scoped rules
Agent Policies
Per-agent rules
Allowed
Needs Approval
Denied
Execute
Run agent, log result
Approval Gate
Wait for human sign-off
Block
Log violation, notify

Policy engine

Four types of policies give you fine-grained control over every aspect of agent behavior.

Model Policies

Restrict which models agents can use. Enforce EU-only providers or cost-tier limits. Set fallback chains.

Tool Policies

Allow or deny specific tools per agent. Control external API access. Require approval for sensitive tools.

Spend Policies

Set daily and monthly caps per project, agent, or organization. Hard or soft limits with configurable alerts.

Data Policies

Define retention periods, export rules, and PII handling per conversation. Automatic redaction and anonymization.

Approval system

Human-in-the-loop approval gates. Require sign-off before agents execute sensitive actions or exceed thresholds.

Approval Triggers

  • Specific tool calls (e.g., create_ticket, send_email)
  • Spend threshold exceeded (configurable per project)
  • Sensitive data access detected
  • First-time agent deployment to production
  • Cross-project or cross-org agent calls

Approval Flows

  • Synchronous: Agent pauses and waits for approval before continuing.
  • Asynchronous: Agent queues the action and notifies approvers via email or Slack.
  • Auto-Approve: Configurable rules for automatic approval of low-risk actions.
  • Role-Based: Different approvers for different action types and risk levels.

Audit trails

Immutable logs of every run. Who, what, when, and which model -- fully searchable and exportable.

Audit Log Last 24 hours
14:32:01 support-agent claude-sonnet-4 search_docs Allowed 342 tokens
14:31:58 support-agent claude-sonnet-4 create_ticket Approved 156 tokens
14:30:22 analytics-bot gpt-4o query_database Allowed 1,204 tokens
14:29:15 onboarding-agent claude-haiku send_email Denied 0 tokens
14:28:03 support-agent claude-sonnet-4 escalate_to_human Approved 89 tokens

Evidence exports

Export full conversation history, model inputs/outputs, and policy decisions for compliance audits and regulatory reviews.

JSON Export

Structured data with full metadata, timestamps, and policy decisions.

CSV Export

Tabular format for spreadsheet analysis and reporting.

Compliance Package

Pre-formatted evidence bundle for AI Act and GDPR audits.

Governance without compromise

Join the waitlist for early access to the most comprehensive AI governance platform in Europe.

Join the Waitlist