Feedmansion Social Media Mcp

by makerprism

2 stars
302 downloads
Not rated
GitHub

About

AI-powered social media scheduler — Connect FeedMansion to Claude, Cursor, ChatGPT, and AI agents via MCP for automated content creation, scheduling, and posting

Details

Author
makerprism
GitHub stars
2
Downloads
302
Categories
Media

- Create content with AI from topics or URLs
- Attach images and videos via URLs, presigned uploads, or base64
- Post to 13 social platforms (Twitter/X, Instagram, Facebook, LinkedIn, Bluesky, TikTok, YouTube, Threads, Mastodon, Pinterest, Reddit, Google Business Profile)
- Human‑in‑the‑loop: nothing goes live without your approval
- 25 MCP tools for drafts, media, scheduling, accounts, and deletion

Setting up with Highlight

This MCP is not yet compatible with Highlight’s one-click setup. However, you can still use it with Highlight by following these steps:

  1. Download and install Highlight from highlightai.com/download
  2. Navigate to the plugins tab and select "Add Custom Plugin"
  3. Configure the plugin with the settings below
    Plugin Name Feedmansion Social Media Mcp
    Command (node, npx, python, etc.)

    Please refer to the README for specific instructions on how to obtain API keys or other required environment variables.

  4. Enable "Start Automatically" if you want the plugin to start when Highlight launches

From the repository

Sign up at feedmansion.com, generate a personal access token at feedmansion.com/settings/tokens, then configure your AI tool using the platform‑specific guides (Claude Code/Desktop, Cursor, Windsurf, OpenClaw, or the generic MCP setup). Once connected, you can ask the agent to create, schedule, or delete content—all changes require your approval in the FeedMansion web UI.

Claude Desktop / Cursor

Paste into your MCP client config file to install this server.

{
    "mcpServers": {
        "feedmansion social media mcp": {
            "feedmansion": {
                "type": "streamable-http",
                "url": "https://feedmansion.com/mcp",
                "headers": {
                    "Authorization": "Bearer fm_pat_YOUR_TOKEN_HERE"
                }
            }
        }
    }
}

McpServers

{
    "feedmansion": {
        "type": "streamable-http",
        "url": "https://feedmansion.com/mcp",
        "headers": {
            "Authorization": "Bearer fm_pat_YOUR_TOKEN_HERE"
        }
    }
}

FeedMansion for AI Agents

Social media automation for AI agents, coding assistants, and automations.

Connect FeedMansion to Claude, Cursor, Windsurf, ChatGPT, or any MCP-compatible AI tool to create content, schedule posts, and manage your social media automatically.

<!--
⚠️ MAINTAINER NOTE: This repo documents FeedMansion's MCP tools.
The canonical tool definitions are in:
feedmansion.com repo → backend/lib/api/mcp_tools.ml

When tools change, update:
- This README.md (tool table)
- platforms/openclaw/SKILL.md (detailed tool docs)
- feedmansion.com/landing-pages/help/en/mcp-setup.md
-->

What You Can Do

- Create content with AI — Generate posts, threads, and captions from topics or URLs
- Attach images and videos — Upload media via URLs or presigned uploads
- Multi-platform posting — One agent, multiple social networks (Twitter/X, Instagram, Facebook, LinkedIn, Bluesky, TikTok, YouTube, Threads, Mastodon, Pinterest, Reddit, Google Business Profile)
- Human oversight — Nothing goes live without your approval in the web UI

Quick Start

1. Get FeedMansion — Sign up at feedmansion.com
2. Generate a token — Create a PAT at feedmansion.com/settings/tokens
3. Connect your AI tool — Choose below:

| AI Tool | Difficulty | Setup |
|---------|------------|-------|
| Claude Code / Claude Desktop | 🟢 Easy | Guide |
| Cursor IDE | 🟢 Easy | Guide |
| Windsurf | 🟢 Easy | Guide |
| OpenClaw | 🟡 Self-hosted | Guide |

Just getting started? We recommend Claude Code — no server setup, works immediately.

<details>
<summary>🔧 Using another MCP-compatible tool?</summary>

See the MCP setup guide for generic configuration that works with any MCP client.
</details>

Example Usage

Once connected, ask your AI:

> "Create a LinkedIn post from this blog article and schedule it for tomorrow morning"

> "Generate a week's worth of tweets from my latest product update"

> "Show me my draft content and queue the approved ones"

> "What's my posting schedule looking like this week?"

How It Works

FeedMansion exposes an MCP (Model Context Protocol) endpoint that gives AI agents access to:

- Your presences (brands/projects)
- Content drafts and posts
- Publishing queues and schedules
- Connected social accounts

AI Agent → MCP → FeedMansion → Social Networks

Approval Flow

Nothing happens without your approval. When agents schedule or delete content, it goes to the Agent Actions page in the FeedMansion web UI. You review and approve before anything goes live.

Available MCP Tools (25 tools)

| Tool | Description |
|------|-------------|
| Presences | |
| list_presences | List all presences (brands/projects) |
| get_presence | Get presence details with accounts, queues |
| update_presence | Request changes to presence settings |
| get_brand_voice | Get full brand voice for external content |
| Drafts | |
| list_drafts | List drafts with optional status filter |
| get_draft | Get a specific draft with details, target accounts, and media |
| create_draft | Create draft from topic, URLs, source items, and/or images |
| edit_draft | Edit draft text or title (returns updated draft with media) |
| batch_create_drafts | Create up to 10 drafts in one call |
| Media | |
| create_media_upload_url | Get presigned URL for large file uploads |
| add_draft_media | Attach media (URL, upload_id, or base64) |
| remove_draft_media | Remove media from a draft |
| Scheduling | |
| schedule_draft | Approve + schedule (queue, specific time, or now) |
| batch_schedule_drafts | Approve + queue multiple drafts |
| add_platforms_to_draft | Add platforms to already-approved content |
| batch_add_platforms_to_drafts | Add platforms to multiple approved drafts (max 25) |
| get_scheduled_and_queued_posts | View scheduled and queued content |
| get_published_posts | View published and failed content |
| list_queue_posts | List all posts in a specific queue |
| list_unscheduled_content | List approved content with no queue/schedule assigned |
| Deletion | |
| request_delete | Request deletion of content or specific posts (requires approval) |
| Accounts | |
| connect_account | Connect a social account (Bluesky, Mastodon) |
| list_unlinked_accounts | List accounts not linked to a presence |
| link_account_to_presence | Link an account to a presence |
| Feedback | |
| submit_feedback | Submit a bug report, feature request, or general feedback |

Key Workflows

Creating a Draft with AI

create_draft accepts multimodal input. Combine any of: topic text, URLs (up to 5), source items (feed/calendar/email), and media.

{
  "name": "create_draft",
  "arguments": {
    "presence_id": "your-presence-id",
    "topic": "Announce our new product feature",
    "urls": ["https://example.com/blog/post"],
    "multi_draft": true
  }
}

Returns draft with content_url and review_url. When multi_draft is true, returns up to 5 draft variations. Always show draft text to user before scheduling.

Attaching Images

Option 1: Public URL (recommended)

The simplest approach. The server downloads the image for you in a single call.

{
  "name": "create_draft",
  "arguments": {
    "presence_id": "...",
    "topic": "...",
    "media_urls": ["https://example.com/image.jpg"]
  }
}

You can also use add_draft_media with media_url to attach media to an existing draft.

Option 2: Presigned URL (for local files)

Use this when you have a local file without a public URL. Requires 2 MCP calls + 1 HTTP PUT.

1. Call create_media_upload_url with the draft's content_id → get upload_url and upload_id
2. PUT the binary file to upload_url with the correct Content-Type header
3. Call add_draft_media with upload_id to finalize

Option 3: Inline base64 (for small files or sandboxed environments)

Pass base64_data + filename via create_draft's media array or add_draft_media. Works well for small images. For larger files, MCP clients that pass tool arguments via shell commands may hit OS argument length limits. Prefer URL download or presigned upload for files over ~100KB.

Scheduling

{
  "name": "schedule_draft",
  "arguments": {
    "presence_id": "...",
    "draft_id": "...",
    "target_account_ids": ["instagram-id", "twitter-id"],
    "publish_now": true
  }
}

Options: queue_id, scheduled_for (ISO 8601), or publish_now.

Adding Platforms to Existing Content

Already scheduled content to 3 platforms and need a 4th? No need to recreate anything:

{
  "name": "add_platforms_to_draft",
  "arguments": {
    "presence_id": "...",
    "content_id": "...",
    "target_accounts": [
      {"account_id": "new-platform-id", "format": "auto"}
    ]
  }
}

For bulk operations (e.g. adding a platform to 50 scheduled drafts), use batch_add_platforms_to_drafts with up to 25 content_ids per call.

Deleting Content or Posts

Use request_delete to propose deletions. The user approves in the Agent Actions page.

Delete an entire content item (and all its scheduled posts):

{
"name": "request_delete",
"arguments": {
"presence_id": "...",
"content_id": "the-content-id"
}
}

Delete specific platform posts (keep others from the same content):

{
"name": "request_delete",
"arguments": {
"presence_id": "...",
"post_ids": ["post-id-1", "post-id-2"]
}
}

Provide exactly one of content_id or post_ids (not both). Only works for pre-publication content and posts.

Why FeedMansion?

- Built for AI — Native MCP support, designed for agent integration
- Multi-platform — 13 platforms supported
- Human in the loop — Agents can't post without approval
- Brand voice — AI generates content that matches your voice
- Smart scheduling — Queue-based posting with timezone support

Links

- FeedMansion — Sign up and start posting
- Help Center — Documentation
- MCP Setup Guide — Detailed MCP instructions
- GitHub — This repo

Support

- Email: support@feedmansion.com
- Help Center

License

MIT

No reviews yet — be the first

Sign in to leave a review

Use Google, GitHub, or an email account so ratings stay tied to real people.

Email sign in

No reviews posted yet.