Connect to the LuftSocial MCP server
The Model Context Protocol (MCP) is an open standard that lets AI assistants (ChatGPT, Claude Desktop, Codex, Cursor, Windsurf) call external tools in a structured, type-safe way. Instead of writing HTTP code, you describe what you want in plain English and the AI executes the action automatically.
LuftSocial's MCP integration lets supported AI assistants work with your authorized LuftSocial workspace and brand. Depending on approved permissions, an assistant can inspect connected accounts, validate content, work with drafts and schedules, publish to connected social platforms, manage media, view performance insights, and use LuftSocial AI features.
The LuftSocial MCP server exposes 11 tools across 5 categories via Streamable HTTP transport. Connect once and your AI agent can schedule posts, generate captions, and validate media seamlessly.
The assistant receives only the permissions granted during connection. It cannot access a workspace or brand that your LuftSocial account cannot access. Always review destination accounts, content, media, scheduling times, and credit-consuming AI actions before approving an important action.
https://mcp.luftsocial.com/mcpPaste this URL into ChatGPT, Codex, Claude, or Cursor remote MCP settings.
Explore platform connection guides
Step-by-step instructions for connecting each supported AI assistant.
ChatGPTConnecting ChatGPT to LuftSocial
Add the LuftSocial remote MCP app and complete OAuth authorization.
Claude DesktopClaude Desktop Connector Setup
Set up the remote MCP connector inside Anthropic Claude Desktop.
CodexCodex IDE & CLI Configuration
Add LuftSocial MCP server to Codex for automated social workflows.
Cursor IDE Remote MCP Setup
Configure Cursor SSE server setup with instant tool discovery.
Copy starter prompts to test your AI assistant
Paste these directly into ChatGPT, Claude, Codex, or Cursor after connecting.
“Check this launch video for Instagram Reels, TikTok, and YouTube Shorts. If it passes, write a caption for each platform and schedule Tuesday 10 AM in my timezone — list destinations before you schedule.”
Calls get_publishing_capabilities (validate_media), preflight_post and generate_content (caption), then publish_post with mode schedule after confirming destinations.
“Show failed publishes from the last 7 days, explain the errors, and tell me which destinations are eligible to retry. Do not retry until I confirm.”
Calls query_posts (views failed and publish_status). Uses publish_post with mode retry only after confirmation.
“What's our top post this month? Rewrite it for LinkedIn and TikTok in our brand voice, save both as drafts, and do not publish.”
Calls query_analytics (content insights), generate_content (caption or transform_caption), then manage_posts with operation create_draft.
“List everything scheduled this week. Move tomorrow's LinkedIn post to Thursday 9 AM ET and update the caption to mention the webinar.”
Calls query_posts with view scheduled, then manage_posts with operation update_scheduled.
When to use MCP
Both surfaces use the same API key. The difference is how you interact: HTTP requests for code-level integrations, natural-language tools for AI agent workflows.
| REST API | MCP Server | |
|---|---|---|
| Best for | Backend scripts, automation pipelines, n8n workflows, custom server integrations | AI coding assistants — Cursor, Claude Desktop, Codex, Windsurf |
| How you interact | HTTP requests to typed JSON endpoints (curl, fetch, requests) | Natural-language prompts — the AI calls the right tool automatically |
| Authentication | Authorization: Bearer ls_sk_… — one header, no extras needed | An MCP API key in an Authorization header — or OAuth (paste URL only) |
| Docs & tooling | OpenAPI reference (Scalar) — browse & test endpoints interactively | MCP setup wizard + searchable tool catalog with 11 tools |
| Setup time | ~2 min — create a key, export env var, make your first request | ~3 min — create a key, paste config snippet into your AI client, ask a question |
| Subscription | Pro or higher required | Pro or higher required |
You can use both in the same project — for example, an n8n automation that calls REST endpoints to batch-schedule posts, while a Cursor agent uses MCP to draft captions and run preflight checks interactively.
Supported Workflows & Prerequisites
What your AI assistant can accomplish and key setup requirements.
9 Supported Automation Workflows
- 1Discover accessible brands and connected social accounts.
- 2Read validation rules and supported post types.
- 3Validate and preflight a proposed post.
- 4Create, update, and delete drafts.
- 5Schedule, update, cancel, publish, and retry eligible posts.
- 6Review publishing state and failed destinations.
- 7Upload, list, sign, transform, and delete media.
- 8Generate captions, hashtags, content ideas, weekly plans, and visual-direction suggestions.
- 9Read top posts, trend signals, and recommended posting times.
Account & Plan Prerequisites
- Active LuftSocial account with access to an active workspace and brand.
- Plan tier with Agent API / MCP access enabled.
- At least one connected social account for publishing workflows (reading validation/rules does not require connected accounts).
- Appropriate workspace/brand permissions & OAuth scopes approved.
- Sufficient workspace AI credits for AI-powered tools.
- A compatible ChatGPT, Codex, or Claude plan/client.
Supported Social Platforms Matrix
Platform capabilities dependent on connected provider accounts and granted scopes.
| Platform | Supported Post Types | Publish | Schedule | Media Transform |
|---|---|---|---|---|
 Facebook | POSTREELSSTORIES | Yes | Yes | No |
 Instagram | POSTREELSSTORIES | Yes | Yes | Yes |
 LinkedIn | POST | Yes | Yes | No |
 TikTok | POST | Yes | Yes | Yes |
 X / Twitter | POSTTHREAD | Yes | Yes | No |
 YouTube | POSTSHORTS | Yes | Yes | No |
 Threads | POSTTHREAD | Yes | Yes | No |
*Platform capabilities depend on the connected provider account, granted provider permissions, media format, LuftSocial plan, and the provider’s current API availability.