@sentinel-atl/adapters

Framework adapters for LangChain, CrewAI, AutoGen, and OpenAI Agents SDK — integrate Sentinel trust in any AI agent framework.

Features

• LangChain — langchainToolWrapper() wraps any LangChain tool with trust verification
• CrewAI — crewaiTaskGuard() guards CrewAI task execution
• AutoGen — createAutoGenMessageFilter() filters multi-agent messages
• OpenAI Agents SDK — openaiAgentGuardrail() adds function-level guardrails
• Universal — withTrust() wraps any async function with trust checks
• Zero dependencies — no framework SDKs required

Install

npm install @sentinel-atl/adapters

Quick Start

import { withTrust, StubTrustVerifier } from '@sentinel-atl/adapters';

const verifier = new StubTrustVerifier({ allowed: true });

const protectedFn = withTrust(verifier, {
  fn: async (input: string) => `Result: ${input}`,
  agentDid: 'did:key:z6Mk...',
  toolName: 'my-tool',
});

const result = await protectedFn('hello');
console.log(result); // "Result: hello"

LangChain

import { langchainToolWrapper } from '@sentinel-atl/adapters';

const wrapped = langchainToolWrapper(verifier, {
  name: 'search',
  description: 'Search the web',
  fn: async (query) => searchWeb(query),
  agentDid: 'did:key:z6Mk...',
});

const result = await wrapped.fn('AI safety');

OpenAI Agents SDK

import { openaiAgentGuardrail } from '@sentinel-atl/adapters';

const guardrail = openaiAgentGuardrail(verifier, {
  agentDid: 'did:key:z6Mk...',
  toolName: 'code_exec',
  fn: async (code: string) => eval(code),
});

API

| Export | Description | |---|---| | withTrust(verifier, config) | Universal trust wrapper for any async function | | langchainToolWrapper(verifier, config) | LangChain tool wrapper | | crewaiTaskGuard(verifier, config) | CrewAI task guard | | createAutoGenMessageFilter(config) | AutoGen message filter | | openaiAgentGuardrail(verifier, config) | OpenAI Agents SDK guardrail | | StubTrustVerifier | Test stub for TrustVerifier interface | | TrustVerifier | Interface for trust verification |

License

MIT