observee 0.1.20

Observee SDK - Tool usage logging, monitoring, authentication, and agent system for LLM integrations

Observee SDK

The complete Observee SDK - All-in-one package for using MCPs with AI agents, authentication management for MCPs, and complete observability for MCP Builders.

Available for both TypeScript/JavaScript and Python.

What's Included

This SDK includes all Observee components:

• 🤖 Agents - MCP tool integration with LLM providers (npm | pypi)
• 🔐 Auth - OAuth authentication for 15+ services (npm | pypi)
• 📊 Logger - Structured logging and monitoring (npm | pypi)

Quick Start

TypeScript/JavaScript

# Install everything at once
npm install @observee/sdk

# Or install individual packages
npm install @observee/agents @observee/auth @observee/logger

Python

# Install individual packages (no meta-package yet)
pip install mcp-agents auth-mcp mcp-logger

# Or install all at once
pip install observee

Usage Examples

TypeScript/JavaScript

Option 1: Import from main package

import { chatWithTools, callMcpAuthLogin, Logger } from "@observee/sdk";

// Use agents
const result = await chatWithTools("Search for news", {
  provider: "anthropic",
  observeeApiKey: "obs_your_key",
});

// Use auth
const authResponse = await callMcpAuthLogin({
  authServer: "gmail",
});

// Use logger
const logger = new Logger({ apiKey: "obs_your_key" });
logger.info("Application started");

Option 2: Import from specific sub-packages

import { chatWithTools } from "@observee/agents";
import { callMcpAuthLogin } from "@observee/auth";
import { Logger } from "@observee/logger";

Python

from observee_agents import chat_with_tools
from observee_auth import call_mcpauth_login

# Use agents
result = chat_with_tools(
    "Search for news",
    provider="anthropic",
    observee_api_key="obs_your_key"
)

# Use auth
auth_response = call_mcpauth_login(auth_server="gmail")
print(f"Login URL: {auth_response['auth_url']}")

Complete Example

TypeScript/JavaScript

import { chatWithTools, callMcpAuthLogin, Logger } from "@observee/sdk";

// Set up logging
const logger = new Logger({
  apiKey: "obs_your_key",
  service: "my-app",
});

async function main() {
  try {
    // Authenticate with Gmail
    logger.info("Starting Gmail authentication");
    const auth = await callMcpAuthLogin({
      authServer: "gmail",
    });
    console.log("Visit:", auth.url);

    // Use AI agent with tools
    logger.info("Starting AI chat with tools");
    const result = await chatWithTools("Search my emails for project updates", {
      provider: "anthropic",
      observeeApiKey: "obs_your_key",
    });

    console.log("AI Response:", result.content);
    logger.info("AI chat completed", {
      toolsUsed: result.toolCalls?.length || 0,
    });
  } catch (error) {
    logger.error("Application error", { error: error.message });
  }
}

main();

Python

from observee_agents import chat_with_tools call_mcpauth_login
import asyncio


async def main():
    try:
        # Authenticate with Gmail
        logger.info("Starting Gmail authentication")
        auth = call_mcpauth_login(auth_server="gmail")
        print(f"Visit: {auth['auth_url']}")

        # Use AI agent with tools
        logger.info("Starting AI chat with tools")
        result = await chat_with_tools(
            "Search my emails for project updates",
            provider="anthropic",
            observee_api_key="obs_your_key"
        )

        print(f"AI Response: {result['content']}")
    except Exception as error:
        logger.error("Application error", {"error": str(error)})

# Run the async function
asyncio.run(main())

Package Documentation

Each component has detailed documentation:

TypeScript/JavaScript

• Agents: README | npm
• Auth: README | npm
• Logger: README | npm

Python

• Agents: README | pypi
• Auth: README | pypi
• Logger: README | pypi

General Documentation

• Observee Docs

Configuration

Environment Variables (Recommended)

Set up your environment variables:

# Core Observee configuration
export OBSERVEE_API_KEY="obs_your_key_here"
export OBSERVEE_CLIENT_ID="your_client_id"

# LLM Provider keys (for agents)
export ANTHROPIC_API_KEY="your_anthropic_key"
export OPENAI_API_KEY="your_openai_key"
export GOOGLE_API_KEY="your_google_key"

When environment variables are set, you can use the SDK without passing API keys to functions:

// No need to pass observeeApiKey when OBSERVEE_API_KEY is set
const result = await chatWithTools("Search for news", {
  provider: "anthropic",
});

const logger = new Logger("my-server", {
  apiKey: process.env.OBSERVEE_API_KEY,
});

Direct Parameter Passing

Alternatively, you can pass API keys directly to functions:

// Pass API key as parameter
const result = await chatWithTools("Search for news", {
  provider: "anthropic",
  observeeApiKey: "obs_your_key",
});

const logger = new Logger("my-server", {
  apiKey: "obs_your_key",
});

Note: API keys passed as parameters take precedence over environment variables.

Features Overview

🤖 AI Agents (@observee/agents)

• Multi-Provider Support: Anthropic Claude, OpenAI GPT, Google Gemini
• MCP Tool Integration: 15+ pre-built tools (Gmail, YouTube, Linear, etc.)
• Smart Filtering: BM25, local embeddings, cloud-based filtering
• Streaming Support: Real-time responses
• Conversation Memory: Persistent chat history

🔐 Authentication (@observee/auth)

• OAuth 2.0 Flows: Gmail, Slack, Notion, Linear, GitHub, and more
• Simple Integration: One-line authentication
• Secure Token Management: Automatic storage and refresh
• Custom Redirects: Support for self-hosted servers

📊 Logging (@observee/logger)

• Structured Logging: JSON-based log format
• Usage Tracking: Monitor API usage and costs
• Multiple Transports: Console, file, and cloud logging
• Performance Monitoring: Track response times and errors

Support

• 📖 Documentation
• 🐛 Issue Tracker
• 💬 Discord Community
• 📧 Email Support

License

All rights reserved. This software is proprietary and confidential.