Stop Your AI Agent Before It Makes a $50,000 Mistake

Your AI agent is about to issue a refund. Should it?

Request Amount ACGP Decision Time
Defective product return $50 OK - Approved 12ms
Suspicious bulk refund $50,000 ESCALATE - Human review 45ms

ACGP checks every agent decision in real-time. Risky actions get caught. Safe ones flow through.

from acgp import GovernanceSteward, CognitiveTrace, PostgresStateStorage

steward = GovernanceSteward.production(
    blueprint_file="blueprint.yaml",
    state_storage=PostgresStateStorage(connection_string="postgresql://runtime/acgp"),
)
trace = CognitiveTrace(
    reasoning="User requested refund for defective item",
    action="issue_refund",
    parameters={"amount": 50}
)
result = steward.evaluate(trace)
print(result.intervention)  # "OK" - approved in 12ms

Try It Now How It Works

Start here: If you are new to ACGP, read ACGP in 10 Minutes first.

For v1.0.0-alpha.2, ACGP has one active external conformance story:

  • Standard is the active claimable profile
  • Regulated Controls Badge is additive
  • Dev Mode is non-conformant

Safety-Critical control material remains in the repository for future-track continuity, but Safety-Critical conformance claims are not available in v1.0.0-alpha.2.

Alpha.2 implementation path

Start with the Standard conformant alpha surface.

Safety-Critical material remains in the repository for future-track continuity and architectural planning, but Safety-Critical conformance claims are not available in v1.0.0-alpha.2.

Choose Your Path

Get your first governed agent running in 5 minutes. Dev Mode Quickstart

Learn how ACGP works — no code required. Concepts Overview

Start with the short mental model, then continue into the full specification set. ACGP in 10 Minutes

What is ACGP?

The Agentic Cognitive Governance Protocol is an open specification for governing autonomous AI agents at runtime. It standardizes how deployments evaluate proposed actions, apply interventions, and preserve audit-visible governance outcomes.

If you want the shortest correct mental model, start with ACGP in 10 Minutes.

If you want the conceptual map, continue with Concepts Overview. How the governance flow works

When to Use ACGP

Perfect for:

  • Autonomous agents that make decisions without constant human oversight
  • High-risk actions like financial transactions, data modifications, or external communications
  • Adaptive AI that learns and changes behavior over time
  • Multi-agent systems that need coordinated governance
  • Compliance requirements that mandate decision audit trails

Not needed for:

  • Simple scripted bots with hard-coded if/else logic
  • Read-only agents that never take actions
  • Sub-10ms latency requirements (ACGP adds 10-500ms depending on Governance Tier)
  • Non-cognitive systems like traditional automation
ACGP vs. Alternatives
Feature Traditional IAM Rate Limiting Monitoring ACGP
Real-time prevention [NO] [YES] [NO] [YES]
Behavioral analysis [NO] [NO] [WARNING] [YES]
Graduated response [NO] [NO] [NO] [YES]
Quality evaluation [NO] [NO] [WARNING] [YES]
Learning/adaptation [NO] [NO] [NO] [YES]
Compliance auditing [WARNING] [NO] [YES] [YES]
Framework agnostic [YES] [YES] [YES] [YES]

Implementation Paths

Choose the implementation path that matches your deployment:

Standard Profile

Standard

Standard-profile alpha governance — ARS scoring, CTQ evaluation, basic tripwires, five primary intervention levels plus orthogonal flag.

~1-2 days to implement

Regulated Controls Badge

Regulated Controls Badge

Additive controls for audit durability, privacy, retention, and evidence integrity on top of the Standard claim surface.

Additive to Standard

Regulated Controls Badge

The active Standard claim can add the **Regulated Controls Badge** for audit-log integrity, data-retention policy, PII redaction, RBAC, and hash-chain evidence.

Compare conformance profiles

Community

GitHub Repository · Join Discord · Follow @acgp_protocol · Contribution Guide

Next Steps