The Problem

AI coding agents have full access to your terminal, filesystem, and secrets. Without governance, an agent can run rm -rf, read .env files, or exfiltrate API keys through tool calls — with no audit trail.

The Solution

pi-governance intercepts every tool call and enforces policy before execution.

pi install npm:@grwnd/pi-governance

What you get immediately:

• Bash blocking — 60+ patterns classify commands as safe/dangerous/needs-review
• DLP — API keys blocked on input, PII masked on output
• RBAC — Role-based tool and path permissions
• Audit — Every decision logged as structured JSON
• HITL — Human approval for sensitive operations
• Budgets — Per-role tool invocation limits
• Config self-protection — Agents cannot modify their own governance files

Customize

Interactive wizard

/governance init

Opens a browser-based wizard to configure roles, DLP, audit, and HITL. Generates YAML config files.

Manual YAML

Create .pi/governance.yaml and governance-rules.yaml — see the Configuration Reference.

Set identity

export PI_GOV_ROLE=project_lead  # analyst | project_lead | admin | auditor
pi
/governance status

Documentation

Full docs at grwnd-ai.github.io/pi-governance.

• Why Governance? — What can go wrong without controls
• Quick Start — Install and configure
• Common Scenarios — Copy-paste configs
• YAML Policies — Full policy reference
• DLP Guide — Data loss prevention
• OpenClaw Integration — MCP tool governance

License

Apache-2.0