Know what your
AI agents actually do
QPThort evaluates every action your AI agent proposes before it executes. Dangerous commands get blocked. Sensitive operations get escalated to you. Every decision is logged to a tamper-proof audit trail.
Works with Claude Code, Cursor, Windsurf, OpenClaw, and any agent with a REST API
Agent attempted:
rm -rf /tmp/production-dataBlocked before execution. Audit record written.
Agent requested:
Read src/components/Button.tsxApproved with signed sanction token. Logged.
How it works
QPThort sits between your AI agent and its actions. One config change, zero code modifications.
Agent proposes action
Before each tool call, your AI agent sends the proposed action to QPThort. This happens automatically via a hook — no agent code changes needed.
Rules evaluate the plan
Constitutional rules check for destructive commands, credential access, data exfiltration, and more. Threat detection classifies every action.
Decision returned instantly
ALLOW, BLOCK, ESCALATE, or MODIFY. Every decision is written to your audit trail before the agent gets a response. Millisecond latency.
Built for trust
Security properties that matter when AI agents have access to your codebase, infrastructure, and data.
Audit-First Architecture
Every decision is written to a hash-chained audit log before the response is returned. If audit is down, all actions are blocked.
Fail-Closed Design
Unknown actions blocked. Timed-out escalations default to block. No silent failures — the system always errs on the side of safety.
Human Escalation
Risky actions route to human principals for review. Approve or reject from your dashboard. Agents cannot resolve their own escalations.
Threat Detection
Detects credential access, data exfiltration, obfuscated commands, shell injection, governance tampering, and 15+ threat categories automatically.
Agent Control
Pause, resume, or revoke any agent from your dashboard in real-time. Per-agent visibility into every decision they've triggered.
5-Minute Setup
One config change to your AI tool. No SDK, no code modifications, no agent rewrites. Works with any tool that supports pre-execution hooks.
Works with your stack
Drop-in integration. No SDK required.
Claude Code
Native hook integration
Cursor
Pre-tool-use hook script
Windsurf
Settings-based governance
OpenClaw
Plugin extension
Any Agent
REST API — any language
Simple pricing
Start free. Scale when you need to.
Free
For hobbyists and personal projects
- ✓ 1,000 decisions/month
- ✓ 3 agents
- ✓ Full audit trail
- ✓ Human escalation
- ✓ Threat detection
Team
For teams shipping with AI agents
- ✓ 25,000 decisions/month
- ✓ Unlimited agents
- ✓ Everything in Free
- ✓ Team members
- ✓ Webhook notifications
Enterprise
For organizations with compliance needs
- ✓ Unlimited decisions
- ✓ Everything in Team
- ✓ Custom rule sets
- ✓ SSO / SAML
- ✓ Dedicated support
Stop hoping your AI agents behave.
Start knowing.
Set up governance in 5 minutes. Free forever for personal use.
Get started for free