arelis-sdk

Arelis AI Governance SDK Plugin for Claude Code

A Claude Code plugin that provides AI governance capabilities through the Arelis AI Governance SDK. Supports both TypeScript (@arelis-ai/ai-governance-sdk) and Python (ai-governance-sdk).

What's included

• AI Governance SDK Skill — Claude automatically assists with governed AI orchestration: createArelis/create_arelis, governedInvoke/governed_invoke, agents.run, governance gates, PII management, policy engines, audit events, risk scoring, compliance proofs, and more.
• Arelis MCP Server — Connects Claude to the Arelis Platform API for live governance operations (policy management, event creation, risk evaluation, proofs, replays).

Prerequisites

• Claude Code v1.0.33 or later
• An Arelis Platform API key (get one at api.arelis.digital)

Installation

From a marketplace

claude plugin install arelis-sdk@<marketplace-name>

Local development

claude --plugin-dir /path/to/ai-governance-plugin

Configuration

Set your Arelis API key as an environment variable before launching Claude Code:

export ARELIS_API_KEY=ak_your_api_key_here

The plugin's MCP server reads ARELIS_API_KEY from the environment. The platform base URL defaults to https://api.arelis.digital.

Usage

Once installed, the plugin activates automatically when Claude detects:

• Imports from @arelis-ai/ai-governance-sdk or from arelis import ...
• References to createArelis, create_arelis, governedInvoke, governed_invoke, withGovernanceGate, with_governance_gate, or GovernanceContext
• Questions about the AI Governance SDK

Skill namespace

All skills are namespaced under arelis-sdk:

/arelis-sdk:ai-governance-sdk

MCP tools

The Arelis MCP server provides tools for:

• AI system registration and management
• Governance policy CRUD and evaluation
• Platform event creation and querying
• Risk evaluation and simulation
• Compliance proof generation and verification
• Causal graph operations
• Quota management

SDK coverage

Feature	TypeScript	Python
createArelis / create_arelis	Yes	Yes
governedInvoke / governed_invoke	Yes	Yes
agents.run (governed agent loop)	Yes	Yes
Governance gates	Yes	Yes
PII scanning and managed config	Yes	Yes
Platform events and risk	Yes	Yes
Policy engine (local + platform)	Yes	Platform only
Causal graphs and proofs	Yes	Yes
MCP, RAG, memory, quotas	Yes	Yes

Project structure

ai-governance-plugin/
├── .claude-plugin/
│   └── plugin.json            # Plugin manifest
├── skills/
│   └── ai-governance-sdk/
│       ├── SKILL.md            # Skill definition
│       ├── references/         # SDK documentation
│       │   ├── typescript/     # TS patterns, API refs, examples
│       │   ├── python/         # Python patterns, API refs, examples
│       │   └── shared/         # Platform API, policies, concepts
│       └── scripts/            # Validation scripts
├── LICENSE
└── README.md

License

MIT — see LICENSE for details.