# How to integrate Facebook MCP with LlamaIndex

```json
{
  "title": "How to integrate Facebook MCP with LlamaIndex",
  "toolkit": "Facebook",
  "toolkit_slug": "facebook",
  "framework": "LlamaIndex",
  "framework_slug": "llama-index",
  "url": "https://composio.dev/toolkits/facebook/framework/llama-index",
  "markdown_url": "https://composio.dev/toolkits/facebook/framework/llama-index.md",
  "updated_at": "2026-05-12T10:11:06.897Z"
}
```

## Introduction

This guide walks you through connecting Facebook to LlamaIndex using the Composio tool router. By the end, you'll have a working Facebook agent that can post new product launch on our page, upload latest event photos to album, reply to comments on latest post through natural language commands.
This guide will help you understand how to give your LlamaIndex agent real control over a Facebook account through Composio's Facebook MCP server.
Before we dive in, let's take a quick look at the key ideas and tools involved.

## Also integrate Facebook with

- [ChatGPT](https://composio.dev/toolkits/facebook/framework/chatgpt)
- [Antigravity](https://composio.dev/toolkits/facebook/framework/antigravity)
- [OpenAI Agents SDK](https://composio.dev/toolkits/facebook/framework/open-ai-agents-sdk)
- [Claude Agent SDK](https://composio.dev/toolkits/facebook/framework/claude-agents-sdk)
- [Claude Code](https://composio.dev/toolkits/facebook/framework/claude-code)
- [Claude Cowork](https://composio.dev/toolkits/facebook/framework/claude-cowork)
- [Codex](https://composio.dev/toolkits/facebook/framework/codex)
- [Cursor](https://composio.dev/toolkits/facebook/framework/cursor)
- [VS Code](https://composio.dev/toolkits/facebook/framework/vscode)
- [OpenCode](https://composio.dev/toolkits/facebook/framework/opencode)
- [OpenClaw](https://composio.dev/toolkits/facebook/framework/openclaw)
- [Hermes](https://composio.dev/toolkits/facebook/framework/hermes-agent)
- [CLI](https://composio.dev/toolkits/facebook/framework/cli)
- [Google ADK](https://composio.dev/toolkits/facebook/framework/google-adk)
- [LangChain](https://composio.dev/toolkits/facebook/framework/langchain)
- [Vercel AI SDK](https://composio.dev/toolkits/facebook/framework/ai-sdk)
- [Mastra AI](https://composio.dev/toolkits/facebook/framework/mastra-ai)
- [CrewAI](https://composio.dev/toolkits/facebook/framework/crew-ai)

## TL;DR

Here's what you'll learn:
- Set your OpenAI and Composio API keys
- Install LlamaIndex and Composio packages
- Create a Composio Tool Router session for Facebook
- Connect LlamaIndex to the Facebook MCP server
- Build a Facebook-powered agent using LlamaIndex
- Interact with Facebook through natural language

## What is LlamaIndex?

LlamaIndex is a data framework for building LLM applications. It provides tools for connecting LLMs to external data sources and services through agents and tools.
Key features include:
- ReAct Agent: Reasoning and acting pattern for tool-using agents
- MCP Tools: Native support for Model Context Protocol
- Context Management: Maintain conversation context across interactions
- Async Support: Built for async/await patterns

## What is the Facebook MCP server, and what's possible with it?

The Facebook MCP server is an implementation of the Model Context Protocol that connects your AI agent and assistants like Claude, Cursor, etc directly to your Facebook Page account. It provides structured and secure access to your Facebook Pages, so your agent can perform actions like publishing posts, managing comments, uploading media, and handling page roles on your behalf.
- Automated content publishing: Have your agent create new posts, photo posts, or video posts directly to your Facebook Page, keeping your audience engaged without manual effort.
- Media management: Effortlessly upload photos to existing albums or create new albums for organized visual storytelling on your Page.
- Interactive engagement: Let your agent add reactions, post comments, or reply to comments, fostering genuine interaction with your followers.
- Page moderation and cleanup: Ask your agent to delete unwanted comments or posts, helping you keep your Facebook Page professional and on-brand.
- Page team management: Assign tasks or roles to users for your Facebook Page, streamlining collaboration and access control.

## Supported Tools

| Tool slug | Name | Description |
|---|---|---|
| `FACEBOOK_ASSIGN_PAGE_TASK` | Assign Page Task | Assigns tasks/roles to a business-scoped user or system user for a specific Facebook Page. Important: This action requires a business-scoped user ID or system user ID from Facebook Business Manager. Regular Facebook user IDs cannot be used. The page must also be managed through Facebook Business Manager for this action to work. Required permissions: business_management, pages_manage_metadata |
| `FACEBOOK_CREATE_COMMENT` | Create Comment | Creates a comment on a Facebook post or replies to an existing comment. |
| `FACEBOOK_CREATE_PHOTO_ALBUM` | Create Photo Album | Creates a new photo album on a Facebook Page. Note: This endpoint requires the 'pages_manage_posts' permission or equivalent permissions to be granted to your Facebook application. This action is publicly visible on the Page; confirm with the user before calling. |
| `FACEBOOK_CREATE_PHOTO_POST` | Create Photo Post | Creates a photo post on a Facebook Page. Requires an image to be provided via either 'url' (publicly accessible image URL) or 'photo' (local image file upload). This action is specifically for posting images with optional captions, not text-only posts. Returns a composite post_id (PageID_PostID); use this for follow-up operations, not the photo/media id alone. |
| `FACEBOOK_CREATE_POST` | Create Post | Creates a new text or link post on a Facebook Page. Requires `pages_manage_posts` permission and manage-level Page role on the target Page. For image posts use FACEBOOK_CREATE_PHOTO_POST; for video posts use FACEBOOK_CREATE_VIDEO_POST — media fields are not supported here. Returns a composite post ID in `PageID_PostID` format, required for FACEBOOK_GET_POST retrieval. |
| `FACEBOOK_CREATE_VIDEO_POST` | Create Video Post | Creates a video post on a Facebook Page. Requires a Page access token with `pages_manage_posts` scope and manage-level permissions on the target page. |
| `FACEBOOK_DELETE_COMMENT` | Delete Comment | Deletes a Facebook comment. Requires a Page Access Token with appropriate permissions for comments on Page-owned content. The page_id parameter helps ensure the correct page token is used for authentication. |
| `FACEBOOK_DELETE_POST` | Delete Post | Permanently deletes a Facebook Page post. Deletion is irreversible — deleted posts cannot be recovered. For bulk deletions, keep throughput to ~1 delete/second to avoid Graph API rate limits. |
| `FACEBOOK_GET_COMMENT` | Get Comment | Retrieves details of a specific Facebook comment. |
| `FACEBOOK_GET_COMMENTS` | Get Comments | Retrieves comments from a Facebook post or comment (for replies). This endpoint requires appropriate permissions: - For page-owned posts: A Page Access Token with 'pages_read_engagement' permission - The API automatically swaps user tokens for page tokens when available API Version: Uses v23.0 which was released May 2025. |
| `FACEBOOK_GET_CONVERSATION_MESSAGES` | Get Conversation Messages | Retrieves messages from a specific conversation. |
| `FACEBOOK_GET_CURRENT_USER` | Get Current User | Validates the access token and retrieves the authenticated user's own profile via /me. Cannot fetch arbitrary users by name or ID. |
| `FACEBOOK_GET_MESSAGE_DETAILS` | Get Message Details | Retrieves details of a specific message sent or received by the Page. |
| `FACEBOOK_GET_PAGE_CONVERSATIONS` | Get Page Conversations | Retrieves a list of conversations between users and the Page. |
| `FACEBOOK_GET_PAGE_DETAILS` | Get Page Details | Fetches details about a specific Facebook Page. |
| `FACEBOOK_GET_PAGE_INSIGHTS` | Get Page Insights | Retrieves analytics and insights for a Facebook Page. Returns metrics like impressions, page views, fan counts, and engagement data. Empty objects (`{}`) in results indicate missing data, not zero values. High-volume calls risk Graph API rate limits (error codes 4/613). |
| `FACEBOOK_GET_PAGE_PHOTOS` | Get Page Photos | Retrieves photos from a Facebook Page. CDN-based URLs (including `source`) are time-limited and expire; download and persist images promptly if long-term access is needed. |
| `FACEBOOK_GET_PAGE_POSTS` | Get Page Posts | Retrieves posts from a Facebook Page. Endpoint choice: Uses /{page_id}/feed instead of /posts or /published_posts because: - /feed returns all content on page timeline (page's posts + visitor posts + tagged posts) - /posts returns only posts created by the page itself - /published_posts returns only published posts by the page (excludes scheduled/unpublished) The /feed endpoint provides the most comprehensive view of page activity. Pagination: follow paging.cursors.after or paging.next across multiple calls until no next cursor exists. Throttling: high-volume pagination can trigger Graph API errors 4 and 613; use backoff between requests. API Version: Uses v23.0 (released May 2025). v20.0 and earlier will be deprecated by Meta. See: https://developers.facebook.com/docs/graph-api/changelog |
| `FACEBOOK_GET_PAGE_ROLES` | Get Page Roles | Retrieves a list of people and their tasks/roles on a Facebook Page. The connected account must have management access to the target Page; otherwise the response may be empty or incomplete. Returned role types include MANAGE and CREATE_CONTENT — verify these before calling tools like FACEBOOK_UPDATE_PAGE_SETTINGS. Recently changed roles may take time to propagate; retry if role data appears stale after an update. |
| `FACEBOOK_GET_PAGE_TAGGED_POSTS` | Get Page Tagged Posts | Retrieves posts where a Facebook Page is tagged or mentioned. Use when monitoring brand mentions or tracking posts that tag your Page but don't appear on your Page's own feed. |
| `FACEBOOK_GET_PAGE_VIDEOS` | Get Page Videos | Retrieves videos from a Facebook Page. |
| `FACEBOOK_GET_POST` | Get Post | Retrieves details of a specific Facebook post. |
| `FACEBOOK_GET_POST_INSIGHTS` | Get Post Insights | Retrieves analytics and insights for a specific Facebook post. Returns metrics like impressions, clicks, and engagement data. Very new posts may return empty metric values; allow a short delay before querying and treat absent fields as partial data. |
| `FACEBOOK_GET_POST_REACTIONS` | Get Post Reactions | Retrieves reactions (like, love, wow, etc.) for a Facebook post. Very recent posts may return empty or partial reactions data; treat missing fields as incomplete coverage, not an error. |
| `FACEBOOK_GET_SCHEDULED_POSTS` | Get Scheduled Posts | Retrieves scheduled and unpublished posts for a Facebook Page. Results are cursor-paginated; follow pagination cursors to retrieve all results beyond the limit. When searching for posts near a specific time, filter to a narrow (~±5 minutes) window. Use this tool to check for existing entries before scheduling new posts to avoid duplicates. |
| `FACEBOOK_ADD_REACTION` | Add Reaction | Adds a LIKE reaction to a Facebook post or comment. Note: Due to API limitations, only LIKE reactions can be added programmatically. This action is user-visible and irreversible — confirm with the user before calling. |
| `FACEBOOK_LIST_MANAGED_PAGES` | List Managed Pages | Retrieves a list of Facebook Pages that the user manages (not personal profiles), including page details, access tokens, and tasks. Requires `pages_show_list` or `pages_read_engagement` OAuth scopes; missing scopes silently return empty results rather than an error. An empty `data` array means the user manages no Pages. Results are paginated via `paging.cursors`; follow `paging.next` until absent to retrieve all Pages when count exceeds `limit`. Graph API throttling (error codes 4, 17, 613) can occur during pagination — use exponential backoff. |
| `FACEBOOK_MARK_MESSAGE_SEEN` | Mark Message Seen | Marks a user's message as seen by the Page, visibly updating the read status in the user's conversation. Note: This action requires an active messaging session with the user. Facebook's messaging policy requires that users have messaged the Page within the last 24 hours for sender actions to work. |
| `FACEBOOK_PUBLISH_SCHEDULED_POST` | Publish Scheduled Post | Publishes a previously scheduled or unpublished Facebook post immediately. This action takes a scheduled or unpublished post and publishes it immediately by setting is_published to true. The post must have been previously created with published=false or with a scheduled_publish_time. Requirements: - The post must exist and be in an unpublished/scheduled state - The user must have admin access to the page that owns the post - The app must have pages_manage_posts permission |
| `FACEBOOK_REMOVE_PAGE_TASK` | Remove Page Task | Removes a user's tasks/access from a specific Facebook Page. Caller must have admin-level rights on the Page. Operates on one page_id at a time; repeat for each page if removing from multiple pages. Partial access may remain if only some tasks are revoked. |
| `FACEBOOK_RESCHEDULE_POST` | Reschedule Post | Changes the scheduled publish time of an unpublished Facebook post. This action updates the scheduled_publish_time of a previously scheduled post. The post must have been created with published=false and a scheduled_publish_time. |
| `FACEBOOK_SEND_MEDIA_MESSAGE` | Send Media Message | Sends a media message (image, video, audio, or file) from the Page to a user. |
| `FACEBOOK_SEND_MESSAGE` | Send Message | Sends a text message from a Facebook Page (not personal profiles) to a user via Messenger. Requires explicit user confirmation before calling, as this action delivers a message to a real end user. |
| `FACEBOOK_TOGGLE_TYPING_INDICATOR` | Toggle Typing Indicator | Shows or hides the typing indicator for a user in Messenger. |
| `FACEBOOK_UNLIKE_POST_OR_COMMENT` | Unlike Post or Comment | Removes a like from a Facebook post or comment. |
| `FACEBOOK_UPDATE_COMMENT` | Update Comment | Updates an existing Facebook comment. IMPORTANT: This action requires a Page Access Token. The comment must belong to a post on a Page that you manage. Use the page_id parameter to ensure the correct page token is used, especially if you manage multiple pages. |
| `FACEBOOK_UPDATE_PAGE_SETTINGS` | Update Page Settings | Updates settings for a specific Facebook Page. Requires the authenticated user to have MANAGE and CREATE_CONTENT tasks for the target page; verify roles via FACEBOOK_GET_PAGE_ROLES. Not all fields (about, description, general_info, etc.) are available for every Page category. |
| `FACEBOOK_UPDATE_POST` | Update Post | Updates an existing Facebook Page post. |
| `FACEBOOK_UPLOAD_PHOTOS_BATCH` | Upload Photos Batch | Uploads multiple photo files in batch to a Facebook Page or Album. Uses Facebook's batch API for efficient multi-photo upload. Maximum 50 photos per batch. |

## Supported Triggers

None listed.

## Creating MCP Server - Stand-alone vs Composio SDK

The Facebook MCP server is an implementation of the Model Context Protocol that connects your AI agent to Facebook. It provides structured and secure access so your agent can perform Facebook operations on your behalf through a secure, permission-based interface.
With Composio's managed implementation, you don't have to create your own developer app. For production, if you're building an end product, we recommend using your own credentials. The managed server helps you prototype fast and go from 0-1 faster.

## Step-by-step Guide

### 1. Prerequisites

Before you begin, make sure you have:
- Python 3.8/Node 16 or higher installed
- A Composio account with the API key
- An OpenAI API key
- A Facebook account and project
- Basic familiarity with async Python/Typescript

### 1. Getting API Keys for OpenAI, Composio, and Facebook

No description provided.

### 2. Installing dependencies

No description provided.
```python
pip install composio-llamaindex llama-index llama-index-llms-openai llama-index-tools-mcp python-dotenv
```

```typescript
npm install @composio/llamaindex @llamaindex/openai @llamaindex/tools @llamaindex/workflow dotenv
```

### 3. Set environment variables

Create a .env file in your project root:
These credentials will be used to:
- Authenticate with OpenAI's GPT-5 model
- Connect to Composio's Tool Router
- Identify your Composio user session for Facebook access
```bash
OPENAI_API_KEY=your-openai-api-key
COMPOSIO_API_KEY=your-composio-api-key
COMPOSIO_USER_ID=your-user-id
```

### 4. Import modules

No description provided.
```python
import asyncio
import os
import dotenv

from composio import Composio
from composio_llamaindex import LlamaIndexProvider
from llama_index.core.agent.workflow import ReActAgent
from llama_index.core.workflow import Context
from llama_index.llms.openai import OpenAI
from llama_index.tools.mcp import BasicMCPClient, McpToolSpec

dotenv.load_dotenv()
```

```typescript
import "dotenv/config";
import readline from "node:readline/promises";
import { stdin as input, stdout as output } from "node:process";

import { Composio } from "@composio/core";

import { mcp } from "@llamaindex/tools";
import { agent as createAgent } from "@llamaindex/workflow";
import { openai } from "@llamaindex/openai";

dotenv.config();
```

### 5. Load environment variables and initialize Composio

No description provided.
```python
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
COMPOSIO_API_KEY = os.getenv("COMPOSIO_API_KEY")
COMPOSIO_USER_ID = os.getenv("COMPOSIO_USER_ID")

if not OPENAI_API_KEY:
    raise ValueError("OPENAI_API_KEY is not set in the environment")
if not COMPOSIO_API_KEY:
    raise ValueError("COMPOSIO_API_KEY is not set in the environment")
if not COMPOSIO_USER_ID:
    raise ValueError("COMPOSIO_USER_ID is not set in the environment")
```

```typescript
const OPENAI_API_KEY = process.env.OPENAI_API_KEY;
const COMPOSIO_API_KEY = process.env.COMPOSIO_API_KEY;
const COMPOSIO_USER_ID = process.env.COMPOSIO_USER_ID;

if (!OPENAI_API_KEY) throw new Error("OPENAI_API_KEY is not set");
if (!COMPOSIO_API_KEY) throw new Error("COMPOSIO_API_KEY is not set");
if (!COMPOSIO_USER_ID) throw new Error("COMPOSIO_USER_ID is not set");
```

### 6. Create a Tool Router session and build the agent function

What's happening here:
- We create a Composio client using your API key and configure it with the LlamaIndex provider
- We then create a tool router MCP session for your user, specifying the toolkits we want to use (in this case, facebook)
- The session returns an MCP HTTP endpoint URL that acts as a gateway to all your configured tools
- LlamaIndex will connect to this endpoint to dynamically discover and use the available Facebook tools.
- The MCP tools are mapped to LlamaIndex-compatible tools and plug them into the Agent.
```python
async def build_agent() -> ReActAgent:
    composio_client = Composio(
        api_key=COMPOSIO_API_KEY,
        provider=LlamaIndexProvider(),
    )

    session = composio_client.create(
        user_id=COMPOSIO_USER_ID,
        toolkits=["facebook"],
    )

    mcp_url = session.mcp.url
    print(f"Composio MCP URL: {mcp_url}")

    mcp_client = BasicMCPClient(mcp_url, headers={"x-api-key": COMPOSIO_API_KEY})
    mcp_tool_spec = McpToolSpec(client=mcp_client)
    tools = await mcp_tool_spec.to_tool_list_async()

    llm = OpenAI(model="gpt-5")

    description = "An agent that uses Composio Tool Router MCP tools to perform Facebook actions."
    system_prompt = """
    You are a helpful assistant connected to Composio Tool Router.
    Use the available tools to answer user queries and perform Facebook actions.
    """
    return ReActAgent(tools=tools, llm=llm, description=description, system_prompt=system_prompt, verbose=True)
```

```typescript
async function buildAgent() {

  console.log(`Initializing Composio client...${COMPOSIO_USER_ID!}...`);
  console.log(`COMPOSIO_USER_ID: ${COMPOSIO_USER_ID!}...`);

  const composio = new Composio({
    apiKey: COMPOSIO_API_KEY,
    provider: new LlamaindexProvider(),
  });

  const session = await composio.create(
    COMPOSIO_USER_ID!,
    {
      toolkits: ["facebook"],
    },
  );

  const mcpUrl = session.mcp.url;
  console.log(`Composio Tool Router MCP URL: ${mcpUrl}`);

  const server = mcp({
    url: mcpUrl,
    clientName: "composio_tool_router_with_llamaindex",
    requestInit: {
      headers: {
        "x-api-key": COMPOSIO_API_KEY!,
      },
    },
    // verbose: true,
  });

  const tools = await server.tools();

  const llm = openai({ apiKey: OPENAI_API_KEY, model: "gpt-5" });

  const agent = createAgent({
    name: "composio_tool_router_with_llamaindex",
        description : "An agent that uses Composio Tool Router MCP tools to perform actions.",
    systemPrompt:
      "You are a helpful assistant connected to Composio Tool Router."+
"Use the available tools to answer user queries and perform Facebook actions." ,
    llm,
    tools,
  });

  return agent;
}
```

### 7. Create an interactive chat loop

No description provided.
```python
async def chat_loop(agent: ReActAgent) -> None:
    ctx = Context(agent)
    print("Type 'quit', 'exit', or Ctrl+C to stop.")

    while True:
        try:
            user_input = input("\nYou: ").strip()
        except (KeyboardInterrupt, EOFError):
            print("\nBye!")
            break

        if not user_input or user_input.lower() in {"quit", "exit"}:
            print("Bye!")
            break

        try:
            print("Agent: ", end="", flush=True)
            handler = agent.run(user_input, ctx=ctx)

            async for event in handler.stream_events():
                # Stream token-by-token from LLM responses
                if hasattr(event, "delta") and event.delta:
                    print(event.delta, end="", flush=True)
                # Show tool calls as they happen
                elif hasattr(event, "tool_name"):
                    print(f"\n[Using tool: {event.tool_name}]", flush=True)

            # Get final response
            response = await handler
            print()  # Newline after streaming
        except KeyboardInterrupt:
            print("\n[Interrupted]")
            continue
        except Exception as e:
            print(f"\nError: {e}")
```

```typescript
async function chatLoop(agent: ReturnType<typeof createAgent>) {
  const rl = readline.createInterface({ input, output });

  console.log("Type 'quit' or 'exit' to stop.");

  while (true) {
    let userInput: string;

    try {
      userInput = (await rl.question("\nYou: ")).trim();
    } catch {
      console.log("\nAgent: Bye!");
      break;
    }

    if (!userInput) {
      continue;
    }

    const lower = userInput.toLowerCase();
    if (lower === "quit" || lower === "exit") {
      console.log("Agent: Bye!");
      break;
    }

    try {
      process.stdout.write("Agent: ");

      const stream = agent.runStream(userInput);
      let finalResult: any = null;

      for await (const event of stream) {
        // The event.data contains the streamed content
        const data: any = event.data;

        // Check for streaming delta content
        if (data?.delta) {
          process.stdout.write(data.delta);
        }

        // Store final result for fallback
        if (data?.result || data?.message) {
          finalResult = data;
        }
      }

      // If no streaming happened, show the final result
      if (finalResult) {
        const answer =
          finalResult.result ??
          finalResult.message?.content ??
          finalResult.message ??
          "";
        if (answer && typeof answer === "string" && !answer.includes("[object")) {
          process.stdout.write(answer);
        }
      }

      console.log(); // New line after streaming completes
    } catch (err: any) {
      console.error("\nAgent error:", err?.message ?? err);
    }
  }

  rl.close();
}
```

### 8. Define the main entry point

What's happening here:
- We're orchestrating the entire application flow
- The agent gets built with proper error handling
- Then we kick off the interactive chat loop so you can start talking to Facebook
```python
async def main() -> None:
    agent = await build_agent()
    await chat_loop(agent)

if __name__ == "__main__":
    # Handle Ctrl+C gracefully
    signal.signal(signal.SIGINT, lambda s, f: (print("\nBye!"), exit(0)))
    try:
        asyncio.run(main())
    except KeyboardInterrupt:
        print("\nBye!")
```

```typescript
async function main() {
  try {
    const agent = await buildAgent();
    await chatLoop(agent);
  } catch (err) {
    console.error("Failed to start agent:", err);
    process.exit(1);
  }
}

main();
```

### 9. Run the agent

When prompted, authenticate and authorise your agent with Facebook, then start asking questions.
```bash
python llamaindex_agent.py
```

```typescript
npx ts-node llamaindex-agent.ts
```

## Complete Code

```python
import asyncio
import os
import signal
import dotenv

from composio import Composio
from composio_llamaindex import LlamaIndexProvider
from llama_index.core.agent.workflow import ReActAgent
from llama_index.core.workflow import Context
from llama_index.llms.openai import OpenAI
from llama_index.tools.mcp import BasicMCPClient, McpToolSpec

dotenv.load_dotenv()

OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
COMPOSIO_API_KEY = os.getenv("COMPOSIO_API_KEY")
COMPOSIO_USER_ID = os.getenv("COMPOSIO_USER_ID")

if not OPENAI_API_KEY:
    raise ValueError("OPENAI_API_KEY is not set")
if not COMPOSIO_API_KEY:
    raise ValueError("COMPOSIO_API_KEY is not set")
if not COMPOSIO_USER_ID:
    raise ValueError("COMPOSIO_USER_ID is not set")

async def build_agent() -> ReActAgent:
    composio_client = Composio(
        api_key=COMPOSIO_API_KEY,
        provider=LlamaIndexProvider(),
    )

    session = composio_client.create(
        user_id=COMPOSIO_USER_ID,
        toolkits=["facebook"],
    )

    mcp_url = session.mcp.url
    print(f"Composio MCP URL: {mcp_url}")

    mcp_client = BasicMCPClient(mcp_url, headers={"x-api-key": COMPOSIO_API_KEY})
    mcp_tool_spec = McpToolSpec(client=mcp_client)
    tools = await mcp_tool_spec.to_tool_list_async()

    llm = OpenAI(model="gpt-5")
    description = "An agent that uses Composio Tool Router MCP tools to perform Facebook actions."
    system_prompt = """
    You are a helpful assistant connected to Composio Tool Router.
    Use the available tools to answer user queries and perform Facebook actions.
    """
    return ReActAgent(
        tools=tools,
        llm=llm,
        description=description,
        system_prompt=system_prompt,
        verbose=True,
    );

async def chat_loop(agent: ReActAgent) -> None:
    ctx = Context(agent)
    print("Type 'quit', 'exit', or Ctrl+C to stop.")

    while True:
        try:
            user_input = input("\nYou: ").strip()
        except (KeyboardInterrupt, EOFError):
            print("\nBye!")
            break

        if not user_input or user_input.lower() in {"quit", "exit"}:
            print("Bye!")
            break

        try:
            print("Agent: ", end="", flush=True)
            handler = agent.run(user_input, ctx=ctx)

            async for event in handler.stream_events():
                # Stream token-by-token from LLM responses
                if hasattr(event, "delta") and event.delta:
                    print(event.delta, end="", flush=True)
                # Show tool calls as they happen
                elif hasattr(event, "tool_name"):
                    print(f"\n[Using tool: {event.tool_name}]", flush=True)

            # Get final response
            response = await handler
            print()  # Newline after streaming
        except KeyboardInterrupt:
            print("\n[Interrupted]")
            continue
        except Exception as e:
            print(f"\nError: {e}")

async def main() -> None:
    agent = await build_agent()
    await chat_loop(agent)

if __name__ == "__main__":
    # Handle Ctrl+C gracefully
    signal.signal(signal.SIGINT, lambda s, f: (print("\nBye!"), exit(0)))
    try:
        asyncio.run(main())
    except KeyboardInterrupt:
        print("\nBye!")
```

```typescript
import "dotenv/config";
import readline from "node:readline/promises";
import { stdin as input, stdout as output } from "node:process";

import { Composio } from "@composio/core";
import { LlamaindexProvider } from "@composio/llamaindex";

import { mcp } from "@llamaindex/tools";
import { agent as createAgent } from "@llamaindex/workflow";
import { openai } from "@llamaindex/openai";

dotenv.config();

const OPENAI_API_KEY = process.env.OPENAI_API_KEY;
const COMPOSIO_API_KEY = process.env.COMPOSIO_API_KEY;
const COMPOSIO_USER_ID = process.env.COMPOSIO_USER_ID;

if (!OPENAI_API_KEY) {
    throw new Error("OPENAI_API_KEY is not set in the environment");
  }
if (!COMPOSIO_API_KEY) {
    throw new Error("COMPOSIO_API_KEY is not set in the environment");
  }
if (!COMPOSIO_USER_ID) {
    throw new Error("COMPOSIO_USER_ID is not set in the environment");
  }

async function buildAgent() {

  console.log(`Initializing Composio client...${COMPOSIO_USER_ID!}...`);
  console.log(`COMPOSIO_USER_ID: ${COMPOSIO_USER_ID!}...`);

  const composio = new Composio({
    apiKey: COMPOSIO_API_KEY,
    provider: new LlamaindexProvider(),
  });

  const session = await composio.create(
    COMPOSIO_USER_ID!,
    {
      toolkits: ["facebook"],
    },
  );

  const mcpUrl = session.mcp.url;
  console.log(`Composio Tool Router MCP URL: ${mcpUrl}`);

  const server = mcp({
    url: mcpUrl,
    clientName: "composio_tool_router_with_llamaindex",
    requestInit: {
      headers: {
        "x-api-key": COMPOSIO_API_KEY!,
      },
    },
    // verbose: true,
  });

  const tools = await server.tools();

  const llm = openai({ apiKey: OPENAI_API_KEY, model: "gpt-5" });

  const agent = createAgent({
    name: "composio_tool_router_with_llamaindex",
    description:
      "An agent that uses Composio Tool Router MCP tools to perform actions.",
    systemPrompt:
      "You are a helpful assistant connected to Composio Tool Router."+
"Use the available tools to answer user queries and perform Facebook actions." ,
    llm,
    tools,
  });

  return agent;
}

async function chatLoop(agent: ReturnType<typeof createAgent>) {
  const rl = readline.createInterface({ input, output });

  console.log("Type 'quit' or 'exit' to stop.");

  while (true) {
    let userInput: string;

    try {
      userInput = (await rl.question("\nYou: ")).trim();
    } catch {
      console.log("\nAgent: Bye!");
      break;
    }

    if (!userInput) {
      continue;
    }

    const lower = userInput.toLowerCase();
    if (lower === "quit" || lower === "exit") {
      console.log("Agent: Bye!");
      break;
    }

    try {
      process.stdout.write("Agent: ");

      const stream = agent.runStream(userInput);
      let finalResult: any = null;

      for await (const event of stream) {
        // The event.data contains the streamed content
        const data: any = event.data;

        // Check for streaming delta content
        if (data?.delta) {
          process.stdout.write(data.delta);
        }

        // Store final result for fallback
        if (data?.result || data?.message) {
          finalResult = data;
        }
      }

      // If no streaming happened, show the final result
      if (finalResult) {
        const answer =
          finalResult.result ??
          finalResult.message?.content ??
          finalResult.message ??
          "";
        if (answer && typeof answer === "string" && !answer.includes("[object")) {
          process.stdout.write(answer);
        }
      }

      console.log(); // New line after streaming completes
    } catch (err: any) {
      console.error("\nAgent error:", err?.message ?? err);
    }
  }

  rl.close();
}

async function main() {
  try {
    const agent = await buildAgent();
    await chatLoop(agent);
  } catch (err: any) {
    console.error("Failed to start agent:", err?.message ?? err);
    process.exit(1);
  }
}

main();
```

## Conclusion

You've successfully connected Facebook to LlamaIndex through Composio's Tool Router MCP layer.
Key takeaways:
- Tool Router dynamically exposes Facebook tools through an MCP endpoint
- LlamaIndex's ReActAgent handles reasoning and orchestration; Composio handles integrations
- The agent becomes more capable without increasing prompt size
- Async Python provides clean, efficient execution of agent workflows
You can easily extend this to other toolkits like Gmail, Notion, Stripe, GitHub, and more by adding them to the toolkits parameter.

## How to build Facebook MCP Agent with another framework

- [ChatGPT](https://composio.dev/toolkits/facebook/framework/chatgpt)
- [Antigravity](https://composio.dev/toolkits/facebook/framework/antigravity)
- [OpenAI Agents SDK](https://composio.dev/toolkits/facebook/framework/open-ai-agents-sdk)
- [Claude Agent SDK](https://composio.dev/toolkits/facebook/framework/claude-agents-sdk)
- [Claude Code](https://composio.dev/toolkits/facebook/framework/claude-code)
- [Claude Cowork](https://composio.dev/toolkits/facebook/framework/claude-cowork)
- [Codex](https://composio.dev/toolkits/facebook/framework/codex)
- [Cursor](https://composio.dev/toolkits/facebook/framework/cursor)
- [VS Code](https://composio.dev/toolkits/facebook/framework/vscode)
- [OpenCode](https://composio.dev/toolkits/facebook/framework/opencode)
- [OpenClaw](https://composio.dev/toolkits/facebook/framework/openclaw)
- [Hermes](https://composio.dev/toolkits/facebook/framework/hermes-agent)
- [CLI](https://composio.dev/toolkits/facebook/framework/cli)
- [Google ADK](https://composio.dev/toolkits/facebook/framework/google-adk)
- [LangChain](https://composio.dev/toolkits/facebook/framework/langchain)
- [Vercel AI SDK](https://composio.dev/toolkits/facebook/framework/ai-sdk)
- [Mastra AI](https://composio.dev/toolkits/facebook/framework/mastra-ai)
- [CrewAI](https://composio.dev/toolkits/facebook/framework/crew-ai)

## Related Toolkits

- [Reddit](https://composio.dev/toolkits/reddit) - Reddit is a social news platform with thriving user-driven communities (subreddits). It's the go-to place for discussion, content sharing, and viral marketing.
- [Linkedin](https://composio.dev/toolkits/linkedin) - LinkedIn is a professional networking platform for connecting, sharing content, and engaging with business opportunities. It's the go-to place for building your professional brand and unlocking new career connections.
- [Active campaign](https://composio.dev/toolkits/active_campaign) - ActiveCampaign is a marketing automation and CRM platform for managing email campaigns, sales pipelines, and customer segmentation. It helps businesses engage customers and drive growth through smart automation and targeted outreach.
- [ActiveTrail](https://composio.dev/toolkits/active_trail) - ActiveTrail is a user-friendly email marketing and automation platform. It helps you reach subscribers and automate campaigns with ease.
- [Ahrefs](https://composio.dev/toolkits/ahrefs) - Ahrefs is an SEO and marketing platform for site audits, keyword research, and competitor insights. It helps you improve search rankings and drive organic traffic.
- [Amcards](https://composio.dev/toolkits/amcards) - AMCards lets you create and mail personalized greeting cards online. Build stronger customer relationships with easy, automated card campaigns.
- [Beamer](https://composio.dev/toolkits/beamer) - Beamer is a news and changelog platform for in-app announcements and feature updates. It helps companies boost user engagement by sharing news where users are most active.
- [Benchmark email](https://composio.dev/toolkits/benchmark_email) - Benchmark Email is a platform for creating, sending, and tracking email campaigns. It's built to help you engage audiences and analyze results—all in one place.
- [Bigmailer](https://composio.dev/toolkits/bigmailer) - BigMailer is an email marketing platform for managing multiple brands with white-labeling and automation. It helps teams streamline campaigns and simplify integration with Amazon SES.
- [Brandfetch](https://composio.dev/toolkits/brandfetch) - Brandfetch is an API that delivers company logos, colors, and visual branding assets. It helps marketers and developers keep brand visuals consistent everywhere.
- [Brevo](https://composio.dev/toolkits/brevo) - Brevo is an all-in-one email and SMS marketing platform for transactional messaging, automation, and CRM. It helps businesses engage customers and streamline communications through powerful campaign tools.
- [Campayn](https://composio.dev/toolkits/campayn) - Campayn is an email marketing platform for creating, sending, and managing campaigns. It helps businesses engage contacts and grow audiences with easy-to-use tools.
- [Cardly](https://composio.dev/toolkits/cardly) - Cardly is a platform for creating and sending personalized direct mail to customers. It helps businesses break through the digital clutter by getting real engagement via physical mailboxes.
- [ClickSend](https://composio.dev/toolkits/clicksend) - ClickSend is a cloud-based SMS and email marketing platform for businesses. It streamlines communication by enabling quick message delivery and contact management.
- [Crustdata](https://composio.dev/toolkits/crustdata) - CrustData is an AI-powered data intelligence platform for real-time company and people data. It helps B2B sales teams, AI SDRs, and investors react to live business signals.
- [Curated](https://composio.dev/toolkits/curated) - Curated is a platform for collecting, curating, and publishing newsletters. It streamlines content aggregation and distribution for creators and teams.
- [Customerio](https://composio.dev/toolkits/customerio) - Customer.io is a customer engagement platform for targeted messaging across email, SMS, and push. Easily automate, segment, and track communications with your audience.
- [Cutt ly](https://composio.dev/toolkits/cutt_ly) - Cutt.ly is a URL shortening service for managing and analyzing links. Streamline your workflows with quick, trackable, and branded short URLs.
- [Demio](https://composio.dev/toolkits/demio) - Demio is webinar software built for marketers, offering both live and automated sessions with interactive features. It helps teams engage audiences and optimize lead generation through detailed analytics.
- [Doppler marketing automation](https://composio.dev/toolkits/doppler_marketing_automation) - Doppler marketing automation is a platform for creating, sending, and tracking email campaigns. It helps you automate marketing workflows and manage subscriber lists for better engagement.

## Frequently Asked Questions

### What are the differences in Tool Router MCP and Facebook MCP?

With a standalone Facebook MCP server, the agents and LLMs can only access a fixed set of Facebook tools tied to that server. However, with the Composio Tool Router, agents can dynamically load tools from Facebook and many other apps based on the task at hand, all through a single MCP endpoint.

### Can I use Tool Router MCP with LlamaIndex?

Yes, you can. LlamaIndex fully supports MCP integration. You get structured tool calling, message history handling, and model orchestration while Tool Router takes care of discovering and serving the right Facebook tools.

### Can I manage the permissions and scopes for Facebook while using Tool Router?

Yes, absolutely. You can configure which Facebook scopes and actions are allowed when connecting your account to Composio. You can also bring your own OAuth credentials or API configuration so you keep full control over what the agent can do.

### How safe is my data with Composio Tool Router?

All sensitive data such as tokens, keys, and configuration is fully encrypted at rest and in transit. Composio is SOC 2 Type 2 compliant and follows strict security practices so your Facebook data and credentials are handled as safely as possible.

---
[See all toolkits](https://composio.dev/toolkits) · [Composio docs](https://docs.composio.dev/llms.txt)