Social Media MCP Server

Social media MCP server for AI clients, posting, and scheduling

Drop the official bundle.social MCP server into Claude Desktop, Claude Code, Cursor or any MCP-aware client and your agent gains first-class tools for posting, scheduling, media uploads and analytics across 14+ platforms.

Last updated: May 2026

View API docs

API key authScheduled publishing14 platforms

Quick start

API request

1// .mcp.json - drop this into Claude Desktop, Claude Code, Cursor or any MCP client2{3  "mcpServers": {4    "bundlesocial": {5      "command": "npx",6      "args": ["-y", "bundlesocial-mcp"],7      "env": { "BUNDLESOCIAL_API_KEY": "sk_live_..." }8    }9  }10}1112// Then, in your AI client:13// "Schedule a LinkedIn + X post for Monday 9am UTC announcing the launch."14// → create_post / schedule_post tools run against bundle.social

Wiring an MCP server to every social platform yourself

Each platform has a different OAuth flow, payload shape and media-upload contract - a lot of glue code to expose as MCP tools.

Rate limits, retries and per-platform edge cases (TikTok privacy, YouTube category IDs, X video duration) leak straight into your tool definitions.

Keeping the MCP tool surface in sync as platforms change APIs is a recurring chore, not a one-time build.

What bundle.social handles

One pre-built MCP server exposes create_post, schedule_post, list_social_accounts, get_analytics and more - across every connected platform.

We absorb platform changes, retries and media pipelines so your tool surface stays stable as platforms churn.

Pairs with the typed REST SDK and the CLI - same auth, same teams, same posts - so agents and humans share state.

Stop hand-rolling MCP tools per platform

Install bundlesocial-mcp once, point it at an API key, and every connected social account becomes a tool your agent can call. The MCP server tracks platform changes so your tool definitions don't have to.

Workflow

How it works

Connect accounts once, then create and schedule posts with channel-specific fields from one API.

Add the server to your MCP client

Drop the JSON snippet into Claude Desktop, Claude Code, Cursor or any MCP-aware client and set BUNDLESOCIAL_API_KEY.

Connect social accounts (once)

OAuth into X, Instagram, TikTok, LinkedIn, YouTube or any of the 14+ platforms from the bundle.social dashboard.

Let your agent call the tools

create_post, schedule_post, upload_media, get_analytics - all available to the agent as first-class MCP tools.

Product architecture

Build social media workflows without rebuilding every platform

Why a social media MCP server matters for agent workflows

A social media MCP server becomes useful when your agent needs to do more than generate copy. Real publishing work includes checking which accounts are connected, choosing the right platform payload, attaching uploaded media, scheduling the post, watching status, and understanding failures. Without a maintained MCP layer, each AI client ends up with a custom tool wrapper around every native platform.

bundle.social gives Claude, Cursor, and other MCP-aware clients one tool surface for social publishing while keeping the platform complexity behind the API. Your agent can call create_post or schedule_post, but your product team does not need to teach it every OAuth flow, media rule, validation message, or retry path separately. That is the practical difference between an agent that can draft a campaign and an agent that can safely operate inside a production workflow.

Model Context Protocol as the stable interface for social tools

The Model Context Protocol is valuable because it gives AI clients a predictable way to discover and call tools. For social media work, that tool surface needs to be conservative, structured, and easy to inspect. A post created by an agent should still have a team, a connected account, a status, platform-specific data, and visible errors that a human operator can review.

That is why the MCP server is only one part of the bundle.social stack. The same underlying workflow is available through the REST API, TypeScript SDK, and CLI, so teams can move between interactive agent usage, backend automation, and human-run scripts without losing the same post objects or account model. MCP becomes the agent-friendly entry point, not a separate social product to maintain.

Build agent publishing without owning every native API edge case

Native platform APIs still matter, and an MCP server should not pretend otherwise. TikTok privacy fields, YouTube video metadata, LinkedIn company pages, X media constraints, Instagram Reels, and Google Business Profile rules all have different requirements. The useful abstraction is not hiding these differences from your product. It is giving the agent one workflow while preserving the fields that matter for each channel.

For teams building AI assistants, internal marketing agents, release-note publishers, or content operations tools, the MCP server keeps the repetitive integration work out of the agent harness. The agent can focus on planning, drafting, selecting accounts, and calling tools, while bundle.social handles account connection, upload references, scheduling state, analytics context, and operational visibility.

Capabilities

What the MCP server exposes

First-party tools for every workflow an agent needs around social publishing - without per-platform glue code.

Posting & scheduling tools

create_post and schedule_post handle one post for many platforms with platform-specific payload blocks. Status comes back structured.

Account & team tools

list_social_accounts, list_teams and switch_team let agents discover which accounts they can publish to before they post.

Media upload tools

upload_media accepts URLs or file paths. Returns an uploadId your agent can reference from create_post - no base64 round-tripping.

Analytics tools

get_post_analytics and get_account_analytics surface per-post and per-account metrics so the agent can close the loop on what it published.

Developer example

Add bundlesocial-mcp to any MCP client

Examples use bundle.social's public API shape: API key authentication, a post date, selected social account types, and platform-specific data.

TypeScript

API request

1// .mcp.json - drop this into Claude Desktop, Claude Code, Cursor or any MCP client2{3  "mcpServers": {4    "bundlesocial": {5      "command": "npx",6      "args": ["-y", "bundlesocial-mcp"],7      "env": { "BUNDLESOCIAL_API_KEY": "sk_live_..." }8    }9  }10}1112// Then, in your AI client:13// "Schedule a LinkedIn + X post for Monday 9am UTC announcing the launch."14// → create_post / schedule_post tools run against bundle.social

Supported content

create_post toolschedule_post toolupload_media toollist_social_accounts toolget_analytics tool

Honest limitations

Requires an MCP-aware client (Claude Desktop, Claude Code, Cursor, or any client implementing the protocol).
Posts created via MCP follow the same per-platform rules as posts created via SDK/API - TikTok still requires privacy, YouTube still has thumbnail rules.
Tool surface is intentionally conservative - destructive actions (delete_account, etc.) are excluded by default; use the SDK if you need them.

Guarantees

Developer-first infrastructure

2% error rate

We handle the platform edge cases, media processing, and rate limits so your requests succeed.

Verbose errors

When native APIs fail, we return human-readable error messages and actionable recovery steps.

Flat pricing

No per-post counting. Predictable pricing for teams managing many users, workspaces, and connected accounts.

Same-day support

Direct access to the engineers building the API. We respond to technical issues the same day. Sometimes the same hour. Test us c;

Resources

Technical guides & documentation

bundlesocial-mcp on npm

Install instructions, tool list and changelog.

Model Context Protocol spec

Anthropic's open protocol for client-tool communication.

FAQ

Questions developers ask before building

Which MCP clients does this work with?

Any client implementing the Model Context Protocol - Claude Desktop, Claude Code, Cursor, Continue, and others. The server is a plain Node process invoked via stdio, so client-specific setup is just where to put the JSON config.

Do I need to write any code?

No. Install with npx, set the API key, and the tools appear in your client. If you want to extend the surface, the same actions are available through the bundle.social SDK and REST API.

How does auth work?

The MCP server reads BUNDLESOCIAL_API_KEY from its env. The API key is org-scoped; the agent can act on any social account connected to that org's teams.

Is this the same as the CLI?

The CLI (bundlesocial-cli) is a shell-out interface - agents that exec subprocesses use it. The MCP server speaks the Model Context Protocol - clients like Claude Desktop and Cursor use it. Both wrap the same REST API.

Social Media CLI

JSON-first CLI for shell-out agents, CI and cron.

Social Media API for AI Agents

The MCP server + CLI + SDK overview.

Claude Code Social Media

Claude-specific setup walkthrough.