@entelog/sdk

Zero-config compliance logging for AI agents. Add a single line of code to capture every agent action with AI-powered summaries, risk scoring, and audit-ready reports.

Install

npm install @entelog/sdk

Quick Start

import { Entelog } from '@entelog/sdk';

const audit = new Entelog('a0_live_xxxxx');

// Log any agent action
await audit.log({
  agent: 'SalesBot',
  action: 'data_access',
  data: {
    query: 'customer lookup',
    customer_id: 'cust_123',
    fields_accessed: ['name', 'email'],
  },
});

// Flush on shutdown
await audit.shutdown();

That's it. Every logged action is automatically:

• Summarized in plain English by AI
• Risk-scored (low / medium / high)
• Visible in your real-time dashboard at entelog.com

Configuration

const audit = new Entelog({
  apiKey: process.env.ENTELOG_API_KEY!,
  batchSize: 50,        // events before auto-flush (default: 50)
  flushInterval: 5000,  // ms between flushes (default: 5000)
  maxRetries: 3,        // retry failed requests (default: 3)
  debug: false,         // enable console logging
});

Action Types

Use any string, but these are the standard categories:

| Action | When to use | |---|---| | data_access | Agent reads from a database, file, or API | | api_call | Agent calls an external service | | generation | Agent generates text, code, or other content | | decision | Agent makes an automated decision | | tool_call | Agent uses a tool or function |

API

| Method | Description | |---|---| | new Entelog(apiKey \| options) | Create a client | | audit.log(event) | Queue an event (auto-flushes when batch is full) | | audit.flush() | Send all queued events immediately | | audit.shutdown() | Stop auto-flush and send remaining events |

Framework Integrations

LangChain

from entelog import Entelog
from langchain.callbacks.base import BaseCallbackHandler

audit = Entelog("a0_live_xxxxx")

class EntelogCallback(BaseCallbackHandler):
    def on_tool_start(self, serialized, input_str, **kwargs):
        audit.log(
            agent="LangChainAgent",
            action="tool_call",
            data={"tool": serialized.get("name"), "input": input_str}
        )

agent = initialize_agent(tools, llm, callbacks=[EntelogCallback()])

OpenAI Function Calling

import { Entelog } from '@entelog/sdk';

const audit = new Entelog('a0_live_xxxxx');

async function executeFunction(name: string, args: any) {
  const result = await functions[name](args);

  await audit.log({
    agent: 'AssistantBot',
    action: 'function_call',
    data: { function: name, arguments: args, result },
  });

  return result;
}

Get your API key

1. Sign up at entelog.com
2. Create a workspace
3. Click Generate Key
4. Use the key here

License

MIT — entelog.com