Learn how to build a powerful AI agent that can analyze social media data using LangChain and SocialAPIs.

Introduction

AI agents are transforming how we interact with data. In this tutorial, we'll build an agent that can:

Fetch social media metrics on demand
Analyze engagement patterns
Generate insights and recommendations
Answer questions about social media performance

Prerequisites

Before we start, make sure you have:

Python 3.9+
A SocialAPIs API key (get one free)
Basic knowledge of Python and APIs

Architecture Overview

Our AI agent will use a simple but powerful architecture:

User Input → Natural language query
LangChain Agent → Interprets the query and decides which tools to use
SocialAPIs Tools → Fetch data from social media platforms
Response Generation → Format and present the results

Step 1: Install Dependencies

bash

1pip install langchain openai requests python-dotenv

Step 2: Set Up Environment

Create a

bash

1.env

file:

bash

1OPENAI_API_KEY=your_openai_key
2SOCIALAPIS_API_KEY=your_socialapis_key

Step 3: Create the SocialAPIs Tool

python

1import os
2import requests
3from langchain.tools import BaseTool
4from dotenv import load_dotenv
5
6load_dotenv()
7
8class FacebookPageTool(BaseTool):
9    name = "facebook_page_details"
10    description = "Get details about a Facebook page including followers, likes, and engagement metrics. Input should be a Facebook page URL."
11
12    def _run(self, url: str) -> str:
13        api_key = os.getenv("SOCIALAPIS_API_KEY")
14        response = requests.get(
15            "https://api.socialapis.io/facebook/pages/details",
16            params={"link": url},
17            headers={"x-api-token": api_key}
18        )
19        
20        if response.ok:
21            data = response.json()
22            return f"""
23            Page: {data.get('name')}
24            Followers: {data.get('followers', 0):,}
25            Likes: {data.get('likes', 0):,}
26            Category: {data.get('category', 'N/A')}
27            """
28        return f"Error: {response.text}"
29
30    def _arun(self, url: str):
31        raise NotImplementedError("Async not supported")

Step 4: Create the Agent

python

1from langchain.agents import initialize_agent, AgentType
2from langchain.chat_models import ChatOpenAI
3
4# Initialize the LLM
5llm = ChatOpenAI(temperature=0, model="gpt-4")
6
7# Create tools
8tools = [FacebookPageTool()]
9
10# Initialize agent
11agent = initialize_agent(
12    tools,
13    llm,
14    agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
15    verbose=True
16)

Step 5: Run the Agent

python

1# Ask the agent questions
2response = agent.run("What are the follower counts for Nike's Facebook page?")
3print(response)
4
5# More complex queries
6response = agent.run("Compare the engagement of Nike and Adidas on Facebook")
7print(response)

Example Output

bash

1> Entering new AgentExecutor chain...
2I need to get the Facebook page details for Nike.
3Action: facebook_page_details
4Action Input: https://facebook.com/nike
5Observation: 
6Page: Nike
7Followers: 38,500,000
8Likes: 38,400,000
9Category: Sports & Recreation
10
11Final Answer: Nike's Facebook page has 38.5 million followers and 38.4 million likes.

Adding More Tools

You can extend the agent with more SocialAPIs tools:

python

1class FacebookPostsTool(BaseTool):
2    name = "facebook_page_posts"
3    description = "Get recent posts from a Facebook page. Input should be a Facebook page URL."
4
5    def _run(self, url: str) -> str:
6        api_key = os.getenv("SOCIALAPIS_API_KEY")
7        response = requests.get(
8            "https://api.socialapis.io/facebook/pages/posts",
9            params={"link": url},
10            headers={"x-api-token": api_key}
11        )
12        
13        if response.ok:
14            data = response.json()
15            posts = data.get('posts', [])[:5]
16            result = "Recent posts:\n"
17            for i, post in enumerate(posts, 1):
18                result += f"{i}. {post.get('text', '')[:100]}...\n"
19                result += f"   Likes: {post.get('likes', 0)}, Comments: {post.get('comments', 0)}\n"
20            return result
21        return f"Error: {response.text}"

Using with MCP (Claude Desktop)

For an even simpler setup, use our MCP server with Claude Desktop:

json

1{
2  "mcpServers": {
3    "socialapis": {
4      "command": "npx",
5      "args": ["-y", "@socialapis/mcp", "YOUR_API_KEY"]
6    }
7  }
8}

Then just ask Claude natural language questions about social media!

Conclusion

You've built a powerful AI agent that can:

Understand natural language queries
Fetch real social media data
Provide actionable insights

Next Steps

Add more platforms (Instagram, TikTok)
Implement caching for faster responses
Add memory for context-aware conversations
Deploy as a chatbot or API

Resources

Questions? Join our community or check our documentation.

Step 3: Create the SocialAPIs Tool

python

1import os
2import requests
3from langchain.tools import BaseTool
4from dotenv import load_dotenv
5
6load_dotenv()
7
8class FacebookPageTool(BaseTool):
9    name = "facebook_page_details"
10    description = "Get details about a Facebook page including followers, likes, and engagement metrics. Input should be a Facebook page URL."
11
12    def _run(self, url: str) -> str:
13        api_key = os.getenv("SOCIALAPIS_API_KEY")
14        response = requests.get(
15            "https://api.socialapis.io/facebook/pages/details",
16            params={"link": url},
17            headers={"x-api-token": api_key}
18        )
19        
20        if response.ok:
21            data = response.json()
22            return f"""
23            Page: {data.get('name')}
24            Followers: {data.get('followers', 0):,}
25            Likes: {data.get('likes', 0):,}
26            Category: {data.get('category', 'N/A')}
27            """
28        return f"Error: {response.text}"
29
30    def _arun(self, url: str):
31        raise NotImplementedError("Async not supported")

Step 4: Create the Agent

python

1from langchain.agents import initialize_agent, AgentType
2from langchain.chat_models import ChatOpenAI
3
4# Initialize the LLM
5llm = ChatOpenAI(temperature=0, model="gpt-4")
6
7# Create tools
8tools = [FacebookPageTool()]
9
10# Initialize agent
11agent = initialize_agent(
12    tools,
13    llm,
14    agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
15    verbose=True
16)

1# Ask the agent questions 2response = agent.run("What are the follower counts for Nike's Facebook page?") 3print(response) 4 5# More complex queries 6response = agent.run("Compare the engagement of Nike and Adidas on Facebook") 7print(response)

Example Output

bash

1> Entering new AgentExecutor chain...
2I need to get the Facebook page details for Nike.
3Action: facebook_page_details
4Action Input: https://facebook.com/nike
5Observation: 
6Page: Nike
7Followers: 38,500,000
8Likes: 38,400,000
9Category: Sports & Recreation
10
11Final Answer: Nike's Facebook page has 38.5 million followers and 38.4 million likes.

Adding More Tools

You can extend the agent with more SocialAPIs tools:

python

1class FacebookPostsTool(BaseTool):
2    name = "facebook_page_posts"
3    description = "Get recent posts from a Facebook page. Input should be a Facebook page URL."
4
5    def _run(self, url: str) -> str:
6        api_key = os.getenv("SOCIALAPIS_API_KEY")
7        response = requests.get(
8            "https://api.socialapis.io/facebook/pages/posts",
9            params={"link": url},
10            headers={"x-api-token": api_key}
11        )
12        
13        if response.ok:
14            data = response.json()
15            posts = data.get('posts', [])[:5]
16            result = "Recent posts:\n"
17            for i, post in enumerate(posts, 1):
18                result += f"{i}. {post.get('text', '')[:100]}...\n"
19                result += f"   Likes: {post.get('likes', 0)}, Comments: {post.get('comments', 0)}\n"
20            return result
21        return f"Error: {response.text}"

Build an AI Social Media Agent with LangChain

Introduction

Prerequisites

Architecture Overview

Step 1: Install Dependencies

Step 2: Set Up Environment

Step 3: Create the SocialAPIs Tool

Step 4: Create the Agent

Step 5: Run the Agent

Example Output

Adding More Tools

Using with MCP (Claude Desktop)

Conclusion

Next Steps

Resources

Build an AI Social Media Agent with LangChain

Introduction

Prerequisites

Architecture Overview

Step 1: Install Dependencies

Step 2: Set Up Environment

Step 3: Create the SocialAPIs Tool

Step 4: Create the Agent

Step 5: Run the Agent

Example Output

Adding More Tools

Using with MCP (Claude Desktop)

Conclusion

Next Steps

Resources