# BoundaryAI

> The deterministic, microsecond-class enforcement layer between AI agents and their actions — blocks dangerous tool calls, stops PII leakage, and keeps physical machines inside safe operating limits.

BoundaryAI sits between an AI agent and every action it tries to take (tool call, API request, shell command, motor command) and evaluates that action against policy before it executes. Enforcement is deterministic (same input → same decision), signed (ed25519 audit chain), and fast enough to run in-line: the policy decision measures 137–170µs median and 0.48–0.60ms p99 on a self-hosted release build, with a 4.8ms median full API round trip including the audit write and signed transparency leaf. Designed for teams shipping autonomous agents, LLM copilots, and physical/robotic systems who need provable guardrails, not best-effort prompt guidance.

## Pages

- [Product overview](/product): How the enforcement engine works and what it protects.
- [Digital agents](/product/digital-agent): Guardrails for LLM agents, copilots, and autonomous workflows.
- [Physical machines](/product/physical-machine): Safety enforcement for robots and industrial machines.
- [Data governance](/product/data-governance): PII detection, redaction, and egress control for AI pipelines.
- [Solutions](/solutions): Industry-specific deployments.
- [Finance](/solutions/finance): Wire limits, approval workflows, and audit trails for finance agents.
- [Healthcare](/solutions/healthcare): PHI redaction and HIPAA-aligned enforcement.
- [Government](/solutions/government): Classified data controls and compliance evidence.
- [Manufacturing](/solutions/manufacturing): Machine-safety enforcement for factory-floor automation.
- [Pricing](/pricing): Plans, enforcement limits, and enterprise options.
- [Security](/security): Threat model, cryptographic design, and third-party audits.
- [What is AI agent enforcement?](/what-is-ai-agent-enforcement): Primer on why prompt guidance is not enough.
- [About](/about): Team, mission, and company background.
- [Contact](/contact): Sales, support, and general enquiries.
- [Book a demo](/meet): Live walkthrough with the BoundaryAI team.

## Docs

- [Docs home](/docs): Getting started, concepts, and integration guides.
- [REST API](/docs/api): HTTP endpoints for evaluating actions and managing policies.
- [Python SDK](/docs/python-sdk): Idiomatic client for Python agent frameworks (LangChain, CrewAI, custom).

## Optional

- [Blog](/blog): Engineering notes, threat research, and product updates.
- [Privacy](/privacy)
- [Terms](/terms)