> ## Documentation Index
> Fetch the complete documentation index at: https://docs.observee.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Quickstart - Using MCPs in AI agents

> Get up and running with Observee Agents in under 5 minutes

## Step 1: Get your API Keys and Set up Marketplace

Go to `https://app.observee.ai` and create an API key for your profile. Then click `Connect` and enable servers

### **Get Your API Keys**

1. **Observee API Key**: Sign up at [**observee.ai**](http://observee.ai) → API Keys → Create new key
2. **LLM API Key**: Get from [**Anthropic**](https://console.anthropic.com/), [**OpenAI**](https://platform.openai.com/), or [**Google**](https://aistudio.google.com/)

### **Environment Setup**

Create a `.env` file:

```shellscript theme={null}
# Required
OBSERVEE_API_KEY=obs_your_api_key_here

# Choose one LLM provider
ANTHROPIC_API_KEY=your_key_here
OPENAI_API_KEY=your_key_here
GOOGLE_API_KEY=your_key_here
```

## Step 2: Authenticate & Get Your Client ID

<Note>
  This step is optional for those who are not setting up an OAuth server.
</Note>

If you are using an OAuth server like Gmail, you can authenticate directly from our Auth SDK to get a client ID for accessing tools. Learn more at [Auth SDK Overview](/authsdk/overview).

**3.1 Install the Auth SDK**

<CodeGroup>
  ```bash Python theme={null}
  pip install mcp-agents
  ```

  ```bash TypeScript theme={null}
  npm install @observee/auth
  ```
</CodeGroup>

**3.2 Start the Auth flow for whichever OAuth Server you have enabled**

<CodeGroup>
  ```python Python theme={null}
  from observee_agents import call_mcpauth_login

  # Start authentication flow (optionally provide your own client_id)
  response = call_mcpauth_login(
      auth_server="gmail",
      client_id="<client_id (uuid)>"  # Optional: use your own UUID
  )
  print(f"Visit: {response['url']}")

  # After you visit the URL and authenticate, you'll receive:
  # {"success":true,"client_id":"<client_id (uuid)>"}
  ```

  ```typescript TypeScript theme={null}
  import { callMcpAuthLogin } from "@observee/auth";

  // Start authentication flow (optionally provide your own client_id)
  const response = await callMcpAuthLogin({
      authServer: "gmail",
      clientId: "<client_id (uuid)>"  // Optional: use your own UUID
  });
  console.log(`Visit: ${response.url}`);

  // After you visit the URL and authenticate, you'll receive:
  // {"success":true,"client_id":"<client_id (uuid)>"}
  ```
</CodeGroup>

<Note>
  This will open a browser window for OAuth authentication. After successful login, you'll receive a `client_id` that you'll use in all subsequent API calls. You can optionally provide your own UUID as the client\_id.

  **Tip:** You can reuse the same client\_id to authenticate multiple services. Each service you authenticate adds its tools to that client\_id's available toolset.
</Note>

## Step 3: Install the Agents SDK

<CodeGroup>
  ```bash Python theme={null}
  pip install mcp-agents
  ```

  ```bash TypeScript theme={null}
  npm install @observee/agents
  ```
</CodeGroup>

### Your First Tool-Powered Conversation

<CodeGroup>
  ```python Python theme={null}
  from observee_agents import chat_with_tools

  # Use the client_id from authentication
  result = chat_with_tools(
      message="Search for recent news about AI developments",
      provider="anthropic",
      client_id="<client_id (uuid)>"  # Use the client_id from Step 2
  )

  print("🤖 AI Response:", result["content"])
  print("🔧 Tools Used:", len(result["tool_calls"]))
  ```

  ```typescript TypeScript theme={null}
  import { chatWithTools } from "@observee/agents";

  // Use the client_id from authentication
  const result = await chatWithTools("Search for recent news about AI developments", {
      provider: "anthropic",
      clientId: "<client_id (uuid)>"  // Use the client_id from Step 2
  });

  console.log("🤖 AI Response:", result.content);
  ```
</CodeGroup>

### Streaming Responses

<CodeGroup>
  ```python Python theme={null}
  import asyncio
  from observee_agents import chat_with_tools_stream

  async def stream_example():
      async for chunk in chat_with_tools_stream(
          message="Check my emails and summarize them",
          provider="anthropic"
      ):
          if chunk["type"] == "content":
              print(chunk["content"], end="", flush=True)

  asyncio.run(stream_example())
  ```

  ```typescript TypeScript theme={null}
  import { chatWithToolsStream } from "@observee/agents";

  for await (const chunk of chatWithToolsStream("Check my emails and summarize them", {
      provider: "anthropic"
  })) {
      if (chunk.type === "content") {
          process.stdout.write(chunk.content);
      }
  }
  ```
</CodeGroup>

### Conversation with Memory

<CodeGroup>
  ```python Python theme={null}
  # First message
  result1 = chat_with_tools(
      message="Search for emails about the product launch",
      client_id="<client_id (uuid)>",
      session_id="assistant_001"
  )

  # Follow-up (remembers context)
  result2 = chat_with_tools(
      message="What date was mentioned in those emails?",
      client_id="<client_id (uuid)>",
      session_id="assistant_001"  # Same session = memory
  )
  ```
</CodeGroup>

🎉 **Congratulations!** You can now:

1. **Connect to 1000+ tools** - Gmail, YouTube, Linear, Slack, and more
2. **Use multiple LLM providers** - Anthropic, OpenAI, Gemini and Groq
3. **Created conversations with memory** - Context preserved across messages

## Next Steps

* [**Examples**](/examples/python/basic-usage) - More code examples
* [**Agents SDK**](/agentsdk/overview) - Complete SDK reference
* [**Tool Management**](/agentsdk/tools) - Work with tools directly
* [**Configuration**](/agentsdk/configuration) - Advanced settings

## Need Help?

* 💬 [Discord Community](https://discord.gg/jnf8yHWJ)
* 📧 [Email Support](mailto:contact@observee.ai)