Skip to main content
Autonomous Agent Security

Give autonomous work clear boundaries—and a stop button

Choose what each Charigent may do. Set spending limits, require a person to review key steps, and stop all work at any time. These controls lower risk, but no system can remove every risk.

Start Free Trial View Pricing
14-day free trialCard requiredCore safety on all plans
The Autonomy Settings screen, showing Full Autonomy, Supervised, and Notification Only consent levels a Charigent can be assigned.

What you get

Layered controls for autonomous workflows

Independent checks reduce the chance that one failed or misconfigured layer defeats every control, though no system can eliminate all risk.

Granular Consent System

Choose from three levels: alerts only, human approval, or work that can run within your rules. You decide exactly what each Charigent may do without approval.

Emergency Kill Switch

Stop new automatic work for everyone, one user, or one Charigent. Work already sent to an outside service may still need to be checked.

Six checks before work runs

Charigent checks whether work is allowed, within budget, happening at a safe pace, appropriate, and ready for review. A person must approve it when your rules require that.

Review & Approval Queue

High-risk or out-of-policy actions pause and queue for human review before executing. Reviewers approve, reject, or escalate with full audit context.

Charigent Heartbeat Monitoring

Configure alerts for silence, response thresholds, or anomalous output patterns. Detection and delivery timing can vary.

Budget Limits & Rate Controls

Set per-action, daily, and monthly spending caps per Charigent. Rate limit action frequency to prevent runaway loops from draining your credits.

Safety checks

Six checks for work that runs on its own

  • Kill Switch
  • Consent Check
  • Budget Guard
  • Rate Limiter
  • Content Moderation
  • Review Approval

Use cases

Built for teams that need constrained, reviewable AI operations

AI controls for larger organizations

Use consent records, audit logs, approval controls, and kill switches as evidence during enterprise security and procurement review. Final acceptance depends on each customer’s requirements.

Regulated Industries

Finance, healthcare, and legal teams can require human review and keep safety and activity records. Each customer must decide whether the setup meets its legal duties.

Customer-facing Charigents

Automatic actions can check permission, safety rules, history, and human approval before they run. Match the controls to how sensitive each customer task is.

See it in action

Current Charigent product views

The Phone Setup screen's consent step, requiring explicit acknowledgement of SMS and voice responsibilities before use.
Consent and compliance checks apply across channels, not just autonomy.
The Deploy screen's Permissions tab, alongside Channels, Personality, Knowledge, and Human handoff for a Charigent.
Permissions are set per Charigent, right next to where you deploy it.

Pricing

Core safety features on all plans. Advanced controls - more than one approval step, activity alerts, and spending rules - on Business ($99/mo).

See full pricing comparison

FAQ

Frequently asked questions

What control do I have over what an autonomous Charigent can do?
Three consent levels: alerts only, human approval required for each action, or work that runs within limits you set. You choose the level per Charigent, and nothing runs outside it.
What happens if I need to stop everything immediately?
The kill switch can stop new automatic work at three scopes -- everyone, one user, or one Charigent. Work already sent to an outside service before the stop may still need to be checked separately.
What checks run before an autonomous action executes?
Six: kill switch status, consent check, budget guard, rate limiter, content moderation, and review approval. Any one of them can block the action.
Which controls are Business-only?
Every plan gets the same safety controls: consent levels, the kill switch, the validation checks each autonomous action passes, and per-agent spending limits. None of those are paywalled. What Business ($99/mo) adds is the ability to connect your own AI provider keys, API access, and custom branding.
Can this guarantee an autonomous Charigent never makes a mistake?
No system can eliminate every risk. These layered controls are designed to lower risk and give you a way to catch and stop problems, not to make autonomous action risk-free.

Put it to work with clear controls and review

Start your 14-day free trial. Card required to start; converts automatically to a paid plan when it ends, unless you cancel first.

Get Started Free