APAAI Documentation

The accountability layer for autonomous AI systems

Core Concepts

Actions

Every operation your AI agent performs is an action. Track them all: database queries, API calls, file operations, emails, and more.

client.actions.create(...)

Policies

Define rules for which actions need approval. Use enforce mode for manual approval or observe mode for monitoring only.

mode: 'enforce' | 'observe'

Evidence

Submit proof that checks were performed: security scans, content reviews, compliance checks, or custom validations.

client.evidence.submit(...)

Approvals

Review and approve actions that match enforcement policies through the dashboard or API.

/approvals/pending

SDKs

📦

TypeScript/JavaScript

npm install apaai-ts-sdk
🐍

Python

pip install apaai
🌐

REST API

Use any HTTP client