MCP for Social Media: How AI Agents Publish Content

What is MCP?

Model Context Protocol is an open standard created by Anthropic that defines how AI models communicate with external tools. Instead of an agent writing code and executing it, MCP lets the agent call pre-defined tools directly — the protocol handles the transport, schema validation, and response formatting.

Think of it like this: before MCP, giving an AI agent access to your database meant it would write SQL, you'd review it, run it, paste the results back. With MCP, the agent calls a query_database tool directly and gets structured results back in the same context window.

Why MCP changes social media publishing

The typical publishing workflow requires context-switching: you write content in one tool, export it, log into a scheduling tool, fill in forms, choose platforms, publish. Every step is manual and disconnected.

With an MCP connection to CodivUpload, publishing becomes a side effect of work you're already doing in your AI context:

Setting up CodivUpload MCP in 30 seconds

Install the codivupload-mcp package via npx. Add it to your MCP client's config with your API key:

{
  "mcpServers": {
    "codivupload": {
      "command": "npx",
      "args": ["-y", "codivupload-mcp"],
      "env": {
        "CODIVUPLOAD_API_KEY": "cdv_your_api_key"
      }
    }
  }
}

Available MCP tools

The CodivUpload MCP server exposes four tools. The agent sees their schemas and knows which parameters to pass based on natural language context.

What MCP Tools Does CodivUpload Expose?

The CodivUpload MCP server exposes five tools. Each one maps to a specific API capability, but the agent handles parameter formatting, validation, and error recovery automatically. Here is the full reference:

When an MCP client connects, it receives these tool schemas. The agent reads the parameter types, required fields, and descriptions — then maps your natural language instruction to the correct tool call without you writing a single line of JSON.

Real Agent Workflows

Abstract tool lists only tell half the story. Here are three concrete agent conversations — the exact back-and-forth between you, the agent, and the CodivUpload MCP server. These are not hypothetical; they reflect actual tool calls the MCP server handles.

publish_post({
  profile_name: "codivupload",
  platforms: ["tiktok", "instagram"],
  post_type: "reel",
  media_urls: ["https://cdn.codivupload.com/u/abc/reel.mp4"],
  description: "New feature just dropped — link in bio for early access.",
  tiktok_description: "POV: shipping at 2am again",
  tiktok_privacy_level: 1,
  instagram_media_type: "REELS"
})

Published to 2/2 platforms. TikTok: friends-only, Instagram: Reels. Post ID: p_8f3a2b.

schedule_post({
  profile_name: "codivupload",
  platforms: ["linkedin", "twitter"],
  post_type: "post",
  description: "We just published our guide to MCP automation. Read it here: https://codivupload.com/blog/mcp-social-media",
  scheduled_date: "2026-04-02T14:00:00Z"
})

Scheduled for Apr 2, 2026 at 2:00 PM UTC. Queued for LinkedIn and X. Post ID: p_1c7e9d.

get_posts({
  limit: 5
})

5 posts returned:
1. "POV: shipping at 2am" — TikTok: published, Instagram: published (2 min ago)
2. "MCP automation guide" — LinkedIn: queued, X: queued (scheduled Apr 2)
3. "Q1 recap thread" — X: published, Threads: published (yesterday)
4. "New pricing tiers" — all 4 platforms: published (Mar 28)
5. "Behind the scenes" — TikTok: failed (token expired), Instagram: published (Mar 25)

MCP vs REST API vs Dashboard

CodivUpload offers three ways to publish: the visual dashboard, the REST API, and the MCP server. They all hit the same backend — the difference is how you interact with it. Pick the one that fits how you already work.

Setting Up MCP in 2 Minutes

Every MCP client stores server configurations in a JSON file. You add CodivUpload once, and the tools become available in every conversation. Here is the setup for the two most popular clients.

{
  "mcpServers": {
    "codivupload": {
      "command": "npx",
      "args": ["-y", "codivupload-mcp"],
      "env": {
        "CODIVUPLOAD_API_KEY": "cdv_your_api_key"
      }
    }
  }
}

{
  "mcpServers": {
    "codivupload": {
      "command": "npx",
      "args": ["-y", "codivupload-mcp"],
      "env": {
        "CODIVUPLOAD_API_KEY": "cdv_your_api_key"
      }
    }
  }
}

What agents can do with these tools

These are real examples of prompts that an MCP-connected agent handles end-to-end:

Why this matters for automation

REST APIs require you to build a client. Webhooks require you to build a listener. MCP requires none of that — you write a prompt and the agent handles parameter mapping, retries, and error handling.

Further reading

MCP is one piece of the CodivUpload automation story. These guides cover the rest: