MCP server · CLI · TypeScript SDK

The social media API for AI agents that can post, schedule, and report.

One API for 14+ platforms - with an official MCP server, a JSON-first CLI, and a fully-typed SDK. Build your own scheduler, vertical product or agent harness on top of bundle.social. Usage-based pricing, EU-hosted.

Get an API keyBrowse MCP package
API key authJSON in, JSON out14+ platformsUsage-based pricing
~/your-agent
bundlesocial-cli
$ npx bundlesocial-cli posts:create \$   -c "We just shipped dark mode 🌙" -i x -i bluesky {  "id": "post_3sf9k2",  "status": "SCHEDULED",  "socialAccountTypes": ["TWITTER", "BLUESKY"],  "data": {    "TWITTER": { "text": "We just shipped dark mode 🌙" },    "BLUESKY": { "text": "We just shipped dark mode 🌙" }  }}$

Integration paths

How agents post through bundle.social

Pick the surface that fits your agent or product. They're all thin layers over the same REST API and TypeScript SDK.

Claude Desktop · Claude Code · Cursor

MCP server

Add the bundle.social MCP server to your AI client and it can list integrations, create and schedule posts, upload media and pull analytics - as first-class tools.

MCP setup docs
.mcp.json
MCP server
{
  "mcpServers": {
    "bundlesocial": {
      "command": "npx",
      "args": ["-y", "bundlesocial-mcp"],
      "env": { "BUNDLESOCIAL_API_KEY": "sk_live_..." }
    }
  }
}

# Then, in your AI client:
# "Schedule a LinkedIn + X post for Monday 9am UTC announcing the launch."
# → create_post / schedule_post tools run against bundle.social

Product architecture

Build AI publishing workflows without rebuilding every platform

Social media API for AI agents that need real execution

A social media API for AI agents has to do more than accept generated copy. Agents need a reliable way to discover connected accounts, upload media, create drafts, schedule posts, check status, and understand why something failed. If those pieces are missing, the agent can write a campaign but cannot safely operate the publishing workflow around it.

bundle.social gives agent builders one infrastructure layer for that work. The MCP server fits clients such as Claude, Cursor, and other MCP-aware tools. The CLI fits shell-out agents, cron jobs, and CI workflows. The TypeScript SDK fits teams building their own agent harness or SaaS product. All three surfaces run on the same account, post, upload, analytics, and status model.

MCP, CLI, and SDK are different doors into the same workflow

Agent stacks rarely stay in one shape forever. A prototype may begin with a CLI command, move into an MCP client, then become a backend service that calls the SDK directly. If each step uses a different integration model, the team ends up rewriting account discovery, post creation, scheduling, and error handling every time the agent architecture changes.

The useful version of this stack keeps the transport flexible while the social workflow stays consistent. MCP gives the AI client native tools, the CLI gives non-interactive processes JSON-first commands, and the SDK gives product teams typed backend control. The same connected social accounts and post IDs remain visible across all of them, which makes automation easier to debug and easier to hand back to humans.

Build agent publishing without pretending platforms are identical

AI agents are good at producing platform variants, but the actual platforms still have different rules. Instagram Reels, TikTok videos, LinkedIn company pages, YouTube uploads, X posts, Google Business Profile updates, and Bluesky posts do not share one perfect payload. A serious agent workflow needs an abstraction that keeps those differences explicit instead of flattening them into weak generic text.

bundle.social handles that middle layer. Your agent can plan and draft the content, then create posts with platform-specific data where the platform requires it. Media references, scheduled dates, lifecycle states, readable errors, analytics context, and dashboard visibility stay attached to the workflow. That lets developers build agent products around social media without inheriting every native API problem as permanent roadmap work.

Built on bundle.social

What you can build

We're infrastructure, not an end-user product. Here's the shape of what builders ship on top of us.

Indie SaaSComing soon

Auto-posting product updates

A solo-founder SaaS pipes changelog entries into bundle.social on release - one webhook, fourteen platforms, zero per-platform code.

GitHub release → worker → bundle.social SDK

Vertical schedulerComing soon

“Postiz for podcasters”

A niche scheduler built entirely on bundle.social: their app handles the audience and UX, bundle.social handles every social connection and publish.

Next.js app → bundle.social SDK + webhooks

Agent harnessComing soon

Trend-jacking content engine

An autonomous agent ranks trending topics, drafts posts, and publishes the winners through the bundle.social MCP server.

LLM agent → bundlesocial-mcp → bundle.social

Real case studies coming as builders ship. Want yours here? Tell us what you're building.

Why bundle.social

Why agent builders pick us

One API for 14+ platforms

X, Instagram, TikTok, LinkedIn, YouTube, Facebook, Pinterest, Reddit, Threads, Bluesky, Mastodon, Discord, Slack, Google Business Profile. No per-platform parser code.

TypeScript SDK with full type safety

Generated from the OpenAPI spec - every request and response typed. Build your own scheduler or agent harness on top of it.

Official MCP server + CLI

First-party packages - bundlesocial-mcp and bundlesocial-cli - not community wrappers. The CLI ships with a SKILL.md for OpenClaw-style agents.

Usage-based pricing

Predictable per-post cost, no surprise per-call fees. At Pro you get 10,000 posts/month - that's $0.01 a post.

EU-based, GDPR-clean

Hosted in the EU. Relevant if you (or your customers) are European and care where social data lives.

Built for production

We absorb the platform edge cases, media processing and rate limits, return verbose actionable errors, and run a public status page.

Pricing for agents

Do the math: $0.01 a post

You pay for posts created, not API calls. At Pro ($100/mo) you get 10,000 posts - that's a cent each. At Business ($400/mo), 100,000 posts - under half a cent. Free covers a side project. Enterprise is custom.

Weekend prototype

Up to 20 posts / month

Free

Free covers 20 posts/month - enough to wire things up and demo the agent.

Indie scheduler · 50 users × ~100 posts

~5,000 posts / month

Pro · $100/mo

10,000 posts included - headroom to grow into.

Agent harness pushing content at scale

~30,000 posts / month

Business · $400/mo

100,000 posts included - ~$0.004 per post.

See full pricing

Comparison

How we stack up for programmatic posting

Capabilitybundle.socialPostizAyrshareUpload-Post
Official MCP serverYesYes (postiz-agent)NoNo
Official CLIYes (+ OpenClaw skill)Yes (skill)NoNo
Typed SDKTypeScript, generated from OpenAPIThin clientJS / Python / PHPBasic
Platforms14+~15~13Fewer
Pricing modelUsage-based ($/post)Subscription tiersSubscription + add-onsSubscription
Open sourceNo (managed service)YesNoNo
EU-hostedYesSelf-host optionUSUS

Comparison compiled July 2026. Competitor capabilities change - check their docs for the latest. Want a side-by-side? bundle.social vs Postiz. See also the unified social media API and the social media API for SaaS.

Getting started

From zero to first post in minutes

01

Sign up & get an API key

Create an account and an API key in the dashboard.

Get an API key
02

Pick your integration path

MCP server for AI clients, the CLI for shell-out agents and CI, or the SDK to build your own product.

03

Connect your first social account

OAuth into X, Instagram, TikTok, LinkedIn or any of the 14+ platforms from the dashboard - once.

04

Make your first post

Call create_post / posts:create / postCreate. Minutes, not days.

bundlesocial-mcp bundlesocial-cli bundlesocial (SDK) API reference

FAQ

Questions builders ask

What is the MCP server for?

It exposes bundle.social as Model Context Protocol tools so AI clients like Claude Desktop, Claude Code and Cursor can post, schedule, list and analyze content directly. Install it with npx bundlesocial-mcp and add it to your client config.

How do AI agents use the CLI?

bundlesocial-cli prints JSON to stdout (one object per command) and status text to stderr - easy for an agent to parse. It ships with a SKILL.md so OpenClaw-style runtimes auto-discover it: npx skills add bundleglobal/bundlesocial-cli.

Do I have to use the CLI or MCP server?

No. They're thin wrappers over the bundlesocial TypeScript SDK and the REST API. Use whichever fits - many builders use the SDK directly to build their own scheduler, vertical product or agent harness.

How is pricing calculated for agents?

By posts created, not API calls. Free includes a small monthly allowance; Pro ($100/mo) includes 10,000 posts (~$0.01/post); Business ($400/mo) includes 100,000; Enterprise is custom. No per-call surprises.

Which platforms are supported?

X, Instagram, TikTok, LinkedIn, YouTube, Facebook, Pinterest, Reddit, Threads, Bluesky, Mastodon, Discord, Slack and Google Business Profile - through one API with per-platform options.

Where is data hosted?

In the EU. We're GDPR-aligned, which matters if you or your customers are European.

Last updated: July 2026

Ship your agent's social layer this afternoon

Grab an API key, drop in the MCP server or CLI, connect an account, and post. 14+ platforms, one API, usage-based pricing.

Status page