@mnemom/agent-integrity-protocol

TypeScript SDK for the Agent Integrity Protocol — real-time thinking block analysis for AI agent alignment.

Installation

npm install @mnemom/agent-integrity-protocol

Usage

import {
  checkIntegrity,
  buildSignal,
  AnthropicAdapter,
  WindowManager,
} from '@mnemom/agent-integrity-protocol';

// Extract thinking block from LLM response
const adapter = new AnthropicAdapter();
const thinking = adapter.extract(response);

// Run integrity analysis
const checkpoint = await checkIntegrity({
  thinkingBlock: thinking.content,
  card: alignmentCard,
  config: {
    agentId: 'my-agent',
    analysisLlm: {
      model: 'claude-haiku-4-5-20251001',
      baseUrl: 'https://api.anthropic.com',
      apiKey: process.env.ANTHROPIC_API_KEY,
      maxTokens: 1024,
    },
  },
});

// Act on the verdict
if (checkpoint.verdict === 'clear') {
  proceed();
} else {
  escalate(checkpoint.concerns);
}

API

See the full documentation and specification.

Requirements

• Node.js >= 18.0.0

License

Apache 2.0. See LICENSE for details.