The Control Plane for AI Agents
Ship agents to production knowing every tool call is allowed, denied, or escalated to a human reviewer before it executes — and every decision is on record.
Self-hosted · Your data never leaves your environment · Any framework
This is what ungoverned production agents do.
These are not attack simulations. Documented incidents from real production deployments, 2025–2026.
Agent posted without approval. Sev-1.
An internal AI agent autonomously posted to an engineering forum, leaving sensitive data exposed for two hours.
No policy on autonomous tool calls.
Coding agent deleted production database.
An AI agent wiped an entire database. Its own words: “I destroyed months of work in seconds.”
No deny policy on delete_database.
Agent bypassed peer approval. 13-hour outage.
An AI coding tool resolved a production issue autonomously, without the required approval step.
No human-in-the-loop enforcement.
AIControl puts the checkpoint back. Every tool call allowed, denied, or escalated to a human reviewer. Every decision logged.
Your agents are already in production. Every tool call they make is either authorized — or it isn't.
No enforcement layer
Agents call APIs, write files, and trigger workflows without any governance checkpoint. You find out what happened after the fact — if you find out at all.
No audit evidence
When a regulator, auditor, or board asks what your agents did last quarter, you have nothing to show them. Logs tell you what happened. They don't prove what was authorized.
No human escalation path
When an agent tool call falls outside clear policy — ambiguous parameters, unexpected context, high-stakes action — there is no mechanism to pause, route to a human reviewer, and resume with a decision on record. Exceptions disappear into logs nobody reads.
One endpoint. Universal governance.
Intercept
Agents send every tool call to AIControl before executing. One API endpoint. Works with LangChain, CrewAI, AutoGen, or any MCP-compatible agent.
Evaluate
AIControl evaluates against your policies using Open Policy Agent. Sub-10ms latency. Allow, deny, or escalate to human review.
Log
Every decision written to an immutable audit trail — tool name, parameters, policy matched, decision reason, timestamp.
{ "decision": "allow", "reason": "default_allow", "audit_event_id": "a3f2...", "duration_ms": 7 }
{ "decision": "deny", "reason": "tool_blacklisted", "audit_event_id": "b7e1...", "duration_ms": 6 }
{ "decision": "review", "reason": "requires_human_review", "review_id": "f2a8...", "duration_ms": 8 }
Everything you need to govern agents at enterprise scale.
Policy Engine
Powered by Open Policy Agent — the CNCF standard used by Kubernetes and Terraform. Update policies in milliseconds without a deployment.
Universal Intercept
Framework-agnostic. Works with any agent on any framework — LangChain, CrewAI, AutoGen, MCP-based agents, or custom code. One integration point. No re-platforming.
Immutable Audit Trail
Every intercept produces an audit event regardless of decision. Append-only store. Export for SOC 2, EU AI Act, and internal governance reports.
Human-in-the-Loop
Ambiguous tool calls pause and route to your compliance team via Slack with approve/deny buttons. Every decision recorded with reviewer identity and timestamp.
Agent Registry
Register, approve, and manage every AI agent in your environment. Track tool allowlists, ownership, model version, and lifecycle status.
Self-Hosted
Docker Compose. Runs in your cloud or on-premises. Your audit data never leaves your environment. Up and running in 30 minutes.
Governance in the critical path
Simple, transparent pricing
All plans include full platform access.
Community
forever
No license key required
- OPA policy enforcement
- approved_tools enforcement
- Rate-based policies
- Audit log: 7-day retention
- React dashboard (basic views)
- HITL review queue — in-dashboard only
- Unlimited agents, unlimited policies
Business
+ $15.00 per million intercepts
License key via self-serve (Stripe coming soon)
- Everything in Community, plus:
- 1-year audit log retention
- Slack HITL notifications
- HITL review queue dashboard view
- Priority email support
Enterprise
+ $25.25 per million intercepts
License key via self-serve (Stripe coming soon)
- Everything in Business, plus:
- OPA health-watch observability
- Policy drift detection + warning feed
- Compliance report export (SOC 2, PCI, HIPAA, GLBA)
- SLA (99.9% uptime guarantee)
Estimate your monthly intercepts
Adjust inputs to see your estimated cost.
~150,000 intercepts/month · Business est. $51/mo · Enterprise est. $153/mo
Community is free for all intercept volumes
No per-agent fees. No penalties for deploying more agents.
Built for regulated industries
Financial services, healthcare, and insurance teams have specific audit evidence requirements. AIControl is designed to meet them.
Immutable Audit Trail
Every intercept logged append-only. Cannot be modified or deleted after write.
Audit Evidence on Demand
Immutable intercept logs queryable to your compliance team's exact requirements. Structured export in development.
Human-in-the-Loop Logging
Every escalation logged with reviewer identity, timestamp, and decision note.
Data Residency
Self-hosted deployment. Your audit data never leaves your environment.
Ready to govern your agents?
See AIControl intercept real tool calls, enforce policies, and produce a compliance audit trail — in a 30-minute demo.