Automated policy enforcement for AI agents

Stop bad agent actions before they execute. Prove every decision.

HaltState checks connected tool and API actions before execution. Automate routine decisions, block prohibited actions, resolve exceptions through AI review, and bring humans in when needed—with signed Proof Packs for investigation and audit.

AI-assisted policy setup. Deterministic enforcement. Signed audit evidence.

Create account View Live Control Room

Building an integration? SDK & MCP documentation →

Set your policies. Automate the decisions.

Define your guardrails with AI assistance. Activate them once, then let deterministic policy checks keep routine work moving.

Set your policies
Use the AI Policy Assistant to help draft controls. An authorised operator reviews and activates them.
Enforce automatically
The policy engine checks each connected action before execution. Actions that satisfy the rules proceed; prohibited actions are blocked.
Resolve exceptions
Route eligible exceptions to the council. Bring in a human for unresolved decisions or actions your policy reserves for human approval.

Your team sets the authority. HaltState applies it at runtime.

Refund policy in action

Example policy: automated e-commerce refund governance.

Valid, non-duplicate refunds up to USD 100 are permitted. Duplicate refunds are blocked. Valid, non-duplicate refunds above USD 100 require human approval.

Automatically allowed

Proposed action: A valid USD 80 refund with no duplicate
Policy rule: Within the configured automatic limit
Decision: ALLOW
Execution: Proceeds without manual review.

Automatically blocked

Proposed action: Another attempt to refund the same order
Policy rule: Duplicate refund is prohibited
Decision: DENY
Execution: Blocked before a second refund executes. No manual review.

Human review required

Proposed action: A valid USD 350 refund with no duplicate
Policy rule: This example reserves refunds above USD 100 for human approval
Decision: APPROVAL_REQUIRED
Execution: Held for the authorised reviewer before execution.

QUARANTINE: Stop runaway agents by blocking further governed actions for the selected agent or session.

Purpose-built for the tool and API layer, HaltState works alongside your host, network and access controls to protect business operations.

Enforcement happens before execution. Signed Proof Packs provide verifiable evidence of governed decisions and recorded outcomes for later review.

Explore Signed Proof Packs Architecture & shared responsibility

Live Control Room

See policy turn into action.

Follow a refund agent through live policy checks, council decisions and signed evidence. Explore the latest retail simulation alongside governed agents working on Moltbook.

View Live Control Room

Find what happened. See why it was allowed. Verify the evidence.

Trace a decision through its outcome, inspect the recorded reason, and verify the linked Proof Packs.

Inspect a verified refund

Where do your agents take action?

Security and AI platforms

Apply your policies before connected agents export data, change records or call production APIs.

Explore agent security

Retail and customer operations

Automate permitted refunds and customer workflows while holding exceptions for review.

Explore retail controls

Audit and incident review

Review what was requested, what policy decided and what outcome was reported.

Explore Signed Proof Packs

Building the integration? Integrate an agent.

Product overview

Watch the overview

See how HaltState checks tool calls before execution, keeps compliant work moving, and creates signed evidence for investigation and audit.

Watch the installation walkthrough

Watch HaltState overview on YouTube

Intercept tool calls before execution

Enforce allow, deny, approve, quarantine

Verify signed evidence trace decisions and outcomes with confidence

Redact protect sensitive customer data

From first risky action to active runtime control.

Connect an action that matters. Set its limits. See your controls working.

1Identify one risky action

2Install the SDK

3Let the Policy Assistant draft controls

4Test ALLOW, APPROVAL_REQUIRED and DENY

5Activate when ready

The onboarding wizard guides agent registration, risk selection and AI-assisted policy drafting. Review the controls, test the outcomes and activate enforcement. Start planning now with the zero-install Action Triage, then bring your blueprint into setup.

Govern the actions that create real business risk.

Retail and customer operations agents can do real good and real harm. HaltState governs the business actions that affect your customers, your money, and your data.

refund.create

Refunds and credits

Require approval for high-value refunds. Deny suspicious refund loops. Preserve proof for audit.

payment.authorize

Payments and purchasing

Enforce thresholds, pause unusual spend, and quarantine runaway agents before money moves.

customer.pii.export

Customer data access

Block unnecessary PII exports, require approval for sensitive reads, and redact evidence safely.

customer.email.send

Customer communications

Control outbound emails and SMS, block policy-violating messages, and keep signed decision evidence.

database.write

Production and database writes

Deny destructive operations, freeze writes during incident windows, and capture policy version and actor.

Put policy at the point where agents take action.

Runtime AI agent governance gives your production workflows an enforceable control point. HaltState evaluates connected refunds, payments, data exports, database writes and customer messages against your policies before the tool or API call executes.

HaltState provides this enforcement layer and preserves verifiable evidence of each decision, helping teams support incident response, internal controls, and regulatory alignment.

        ALLOW
        refund.create USD 80
        APPROVAL_REQUIRED
        refund.create USD 350
        DENY
        refund.create duplicate
      

Built for evidence-driven AI governance.

Regulators and enterprise customers increasingly expect more than policy documents. They expect proof that controls operate at runtime. HaltState helps teams preserve action-level evidence for internal audits, incident response, and governance programs aligned to frameworks such as California SB 53 / TFAIA, the EU AI Act, New York RAISE, NIST AI RMF, ISO/IEC 42001, SOC 2 control narratives, and state-level automated decision-making laws.

California SB 53 / TFAIA

Transparency and incident-response evidence for runtime decisions.

EU AI Act

Human oversight, logging, and action-control support.

New York RAISE

Evidence for frontier AI safety and reporting direction.

NIST / ISO / SOC 2

Operational evidence and control narratives.

State ADM laws

Automated decisioning transparency support.

Framework-agnostic by design.

Connect with Python, TypeScript/JavaScript, Go, Java or Rust. Apply the same control pattern to framework tools, custom orchestration and your own backend functions.

Choose your SDK

Investigate from the tools you already use.

Connect your MCP client and ask what an agent did, find customer interactions and verify evidence directly from your assistant.

Connect your MCP client

OpenClaw runtime safety

Put policy between OpenClaw and its tools.

Use pre-execution checks, approvals, scoped stops, and evidence logs before an OpenClaw agent can delete files, send messages, spend money, or change production systems.

Explore HaltState for OpenClaw