One kernel.
Deterministic governance.

PolicyGate is a deterministic policy engine that governs autonomous systems at runtime. Same kernel, two verticals: AI agents and autonomous trading. Every decision logged. Every action governed.

See What We've Shipped

The Problem

Autonomous systems operate with dangerous defaults. Every day.

Publicly bound to 0.0.0.0 with no authentication

Running inside privileged or root containers

API keys sitting in plaintext .env files

Config files world-writable on shared hosts

Trading bots executing without position limits or drawdown controls

Autonomous strategies with no audit trail for regulatory compliance

These are real risks. ClawShield detects agent misconfigurations today. PolicyGate Capital enforces trading constraints at runtime.

How PolicyGate Works

One kernel. Two verticals. Every decision deterministic.

┌─────────────────────────────────────────────────┐
│              PolicyGate Kernel                   │
│   Intent → Policy Evaluation → Decision → Log   │
│   (frozen dataclasses · deterministic replay)    │
└──────────┬──────────────────────┬────────────────┘
           │                      │
 ┌─────────┴─────────┐  ┌────────┴──────────┐
 │  PolicyGate Agent  │  │ PolicyGate Capital │
 │  AI agent tool     │  │ Trading order      │
 │  governance        │  │ governance         │
 └──────────┬─────────┘  └─────────┬──────────┘
           │                      │
 ┌─────────┴──────────────────────┴────────────┐
 │              ClawShield                      │
 │   Static pre-deployment security scanning    │
 └──────────────────────────────────────────────┘

The core primitive: every action is an intent that passes through policy evaluation before execution.

The same kernel powers both verticals — frozen dataclasses, deterministic replay, append-only audit trails.

  • PolicyGate Agent — governs AI agent tool access via YAML policy DSL
  • PolicyGate Capital — governs autonomous trading (position limits, drawdown controls, order validation)
  • ClawShield — static pre-deployment scanning sits below both runtime layers

Shipped

Three packages. All open source. All on PyPI or GitHub.

PolicyGate

Runtime governance for AI agents

93 tests
  • YAML policy DSL
  • Glob matching
  • Rate limits
  • guard() / evaluate() SDK
  • LangChain integration

PolicyGate Capital

Runtime governance for autonomous trading

v0.2.0
  • Position limits
  • Drawdown controls
  • Order validation
  • Execution throttles
  • Replayable audit trails

ClawShield v0.4.1

Static security audit for AI agent deployments

  • 11 detection rules
  • Network exposure
  • Auth & containers
  • Sandbox, shell, browser
  • Secrets, permissions, logging
pip install clawshield

How It Works

1

Scan

ClawShield audits agent configs, containers, secrets, and permissions against security policies before deployment.

2

Enforce

The PolicyGate kernel evaluates every intent against policy rules at runtime — deterministic, no exceptions.

3

Audit

Every decision is logged with full context. Append-only audit trails enable deterministic replay and compliance reporting.

Roadmap

Shipped

PolicyGate kernel

Deterministic policy engine, 93 tests, on GitHub

PolicyGate Capital v0.2.0

Position limits, drawdown, order validation, audit trails, on PyPI

ClawShield v0.4.1

11 detection rules, static security scanning, on PyPI

LangChain integration

guard_tool() wraps tools with policy enforcement

Next

Broker API Integration

Connect PolicyGate Capital to live broker/exchange APIs for production trading governance.

Docker-Aware Scanning

ClawShield detects exposure from docker-compose port mappings, not just agent config.

Additional Runtime Adapters

Support for more AI agent frameworks beyond OpenClaw.

Continuous Monitoring

Watch mode for live detection of configuration drift.

Policy Bundles

Pre-built policy sets for common deployment profiles.

Build With Us

We're looking for a technical cofounder with deep systems engineering experience — event-driven architecture, deterministic runtime design, reliability engineering. If you've built infrastructure where correctness matters, let's talk.

Get in Touch

Stay Updated

Get PolicyGate updates, new releases, and early access.

No spam. Unsubscribe anytime.