Skip to main content
Glama

ReelFarm MCP Server

An MCP (Model Context Protocol) server for the ReelFarm API. Create slideshows, manage automations, publish to TikTok, search Pinterest for images, and more — all from any MCP-compatible AI client.

Tools

25 tools covering the full ReelFarm API:

Category

Tools

Account

get_account

Slideshows

generate_slideshow, create_slideshow, slideshow_status

Automations

create_automation, list_automations, get_automation, update_automation, delete_automation, run_automation

Schedules

add_schedule, update_schedule, delete_schedule

Videos

list_videos, get_video, get_video_analytics, publish_video_via_automation

TikTok

publish_to_tiktok, list_tiktok_accounts, list_tiktok_posts

Collections

list_collections, get_collection_images

Library

list_library_niches, search_library, get_library_profile

Pinterest

search_pinterest

Related MCP server: spoooler

Prerequisites

Quick Start

Add this to your MCP client's config file (replace rf_your_api_key_here with your actual key):

{
  "mcpServers": {
    "reelfarm": {
      "command": "uvx",
      "args": ["reelfarm-mcp"],
      "env": {
        "REELFARM_API_KEY": "rf_your_api_key_here"
      }
    }
  }
}

That's it. The config is the same for every client — the only difference is where the file lives.

Config file locations

Client

Config file

How to open

Claude Desktop

~/Library/Application Support/Claude/claude_desktop_config.json (macOS) · %APPDATA%\Claude\claude_desktop_config.json (Windows)

Claude → Settings → Developer → Edit Config

Cursor

.cursor/mcp.json in your project root

Settings → MCP → Add new MCP server

Windsurf

~/.codeium/windsurf/mcp_config.json

Windsurf Settings → Cascade → MCP → Add Server → Add custom server

VS Code (Copilot)

.vscode/mcp.json

Edit directly (uses "servers" instead of "mcpServers" — see below)

VS Code note: VS Code uses a "servers" key instead of "mcpServers":

{
  "servers": {
    "reelfarm": {
      "command": "uvx",
      "args": ["reelfarm-mcp"],
      "env": {
        "REELFARM_API_KEY": "rf_your_api_key_here"
      }
    }
  }
}

Claude Code (CLI)

claude mcp add reelfarm \
  -e REELFARM_API_KEY=rf_your_api_key_here \
  -- uvx reelfarm-mcp

Verify it's registered:

claude mcp list

Any other MCP client (generic stdio)

The server communicates over stdio. Point your MCP client at:

REELFARM_API_KEY=rf_your_api_key_here uvx reelfarm-mcp

Installation (alternative to uvx)

If you prefer installing the package directly rather than using uvx:

pip install reelfarm-mcp

Then use "command": "reelfarm-mcp" (no args) in any of the configs above instead of uvx.

Using a local clone

If you're developing or want to run from source:

{
  "mcpServers": {
    "reelfarm": {
      "command": "uv",
      "args": [
        "run",
        "--directory", "/absolute/path/to/reelfarm-mcp",
        "reelfarm-mcp"
      ],
      "env": {
        "REELFARM_API_KEY": "rf_your_api_key_here"
      }
    }
  }
}

Or via Claude Code CLI:

claude mcp add reelfarm \
  -e REELFARM_API_KEY=rf_your_api_key_here \
  -- uv run --directory /absolute/path/to/reelfarm-mcp reelfarm-mcp

Usage Examples

Once connected, you can ask your AI assistant things like:

  • "Generate a slideshow about 5 morning habits, casual tone, 6 slides"

  • "List my automations and show which ones are active"

  • "Create an automation that posts fitness slideshows to TikTok every day at 2pm Pacific"

  • "Show me my TikTok post analytics for the last 30 days sorted by views"

  • "Search Pinterest for aesthetic coffee images"

  • "What niches are available in the slideshow library?"

  • "How many credits do I have left?"

Remote deployment (use it from your phone)

The server defaults to stdio. Set MCP_TRANSPORT=streamable-http and it serves over HTTP instead, which is what Claude's custom connectors — including the mobile app — talk to.

Auth. Claude's connector UI takes a URL but has no field for an Authorization header, so the shared secret goes in the URL path. Set MCP_SECRET and the server mounts at /<MCP_SECRET>/mcp; any other path 404s. This is single-user auth: anyone with the full URL can act as your ReelFarm account. Don't paste it into screenshots, issues, or commits.

Generate a secret:

python -c 'import secrets; print(secrets.token_urlsafe(32))'

Deploy to Fly.io

fly.toml and Dockerfile are in the repo. The machine scales to zero and suspends when idle, so it costs well under $1/month and wakes in about a second.

brew install flyctl && fly auth signup
fly launch --no-deploy --copy-config --name YOUR-UNIQUE-APP-NAME
fly secrets set REELFARM_API_KEY=rf_your_key MCP_SECRET=your_generated_secret
fly deploy

Your connector URL is then https://YOUR-APP-NAME.fly.dev/<MCP_SECRET>/mcp.

Verify it before wiring up the client:

curl -sX POST https://YOUR-APP.fly.dev/YOUR-SECRET/mcp -H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

Connect it

On claude.ai: Settings → Connectors → Add custom connector, paste the URL. It syncs to the mobile app automatically.

To rotate the secret, fly secrets set MCP_SECRET=... and update the connector URL — the old path stops resolving on the next deploy.

API Rate Limits

  • 20 requests per 60-second sliding window

  • Max 3 slideshows generating simultaneously

Development

git clone https://github.com/reelfarm/reelfarm-mcp.git
cd reelfarm-mcp
uv sync

# Run the server locally
REELFARM_API_KEY=rf_your_key uv run reelfarm-mcp

Troubleshooting

Problem

Fix

uvx: command not found

Install uv: curl -LsSf https://astral.sh/uv/install.sh | sh

Tools don't appear after config change

Restart the app (Claude Desktop, Cursor, etc.)

REELFARM_API_KEY is not set

Double-check the env block in your config — the key must start with rf_

FastMCP.__init__() got an unexpected keyword argument

Your mcp SDK is outdated or too new. Run uv sync to get compatible versions

Server seems to hang when run directly

This is normal — MCP servers communicate over stdio and produce no terminal output. Your MCP client launches it automatically

License

MIT — see LICENSE.

Available Tools

26 tools
add_scheduleA

Add a new cron schedule to an automation.

Args: automation_id: The automation UUID. cron: Cron expression in Pacific time, e.g. "0 14 * * *" for 2pm daily.

Inputs: cron (required), automation_id (required).

create_automationA

Create a recurring automation that generates and publishes slideshows to TikTok on a cron schedule.

Args: tiktok_account_id: Target TikTok account ID (from list_tiktok_accounts). schedule: Array of cron objects, e.g. [{"cron": "0 14 * * *"}]. Pacific time. title: Descriptive name for this automation. product_id: Link to a product in the products table. slideshow_hooks: Hook/topic templates the AI rotates through. style: Natural-language prompt controlling text items, fonts, sizes, positions, tone. language: Language for generated content (default "English"). tiktok_post_settings: TikTok posting config (caption, visibility, auto_post, etc.). image_settings: Image and display config (aspect_ratio, overlays, collections, etc.).

Inputs: schedule (required), tiktok_account_id (required), style, title, +5 more.

create_slideshowA

Create a slideshow with full manual control over every slide's image, text, and styling.

No AI generation — renders in ~10 seconds.

Args: slides: Array of slide objects. Each slide has image_url or image_urls, text_items, etc. title: Slideshow title (default "API Slideshow"). aspect_ratio: "4:5" (default), "9:16", "1:1", or "16:9". text_position: "top", "center" (default), or "bottom". export_as_video: Export as .mp4 in addition to rendered images. duration: Seconds per slide in video (default 4). Only applies when export_as_video is true. is_bg_overlay_on: Dark overlay on body slides. is_bg_overlay_on_hook_image: Dark overlay on hook slide. background_opacity: Overlay opacity 0-100 (default 20). keep_original_aspect_ratio: Use each image's native aspect ratio.

Inputs: slides (required), title, duration, aspect_ratio, +6 more.

delete_automationA

Permanently delete an automation and all its scheduled jobs.

Args: automation_id: The automation UUID.

Inputs: automation_id (required).

delete_scheduleA

Remove a single schedule job from an automation.

Args: automation_id: The automation UUID. job_id: The job to remove.

Inputs: job_id (required), automation_id (required).

generate_slideshowA

Generate a slideshow from a natural-language prompt. AI picks text, images, and styling.

Rendering is async (~45s). Poll slideshow_status to track progress.

Args: additional_context: Prompt controlling topic, slide count, text items, fonts, tone, etc. images: Optional list of image URLs for slide backgrounds (0-indexed: images[0] → slide 0).

Inputs: additional_context (required), images.

get_accountA

Get account info including subscription tier, remaining credits, and billing cycle reset date.

Inputs: no named parameters.

get_automationA

Get full details of a single automation.

Args: automation_id: The automation UUID.

Inputs: automation_id (required).

get_collection_imagesA

Get images in a collection, paginated.

Args: collection_id: The collection ID. limit: Max results (default 100, max 100). offset: Pagination offset.

Inputs: collection_id (required), limit, offset.

get_library_profileB

Get full details of a slideshow library profile including all slideshow images.

Args: profile_id: The profile ID.

Inputs: profile_id (required).

get_videoB

Get full details for a single video including slideshow images and TikTok publish status.

Args: video_id: The video ID.

Inputs: video_id (required).

get_video_analyticsA

Get TikTok post analytics for a published video (views, likes, comments, shares, bookmarks).

The video must have been published to TikTok.

Args: video_id: The video ID.

Inputs: video_id (required).

list_automationsA

List all automations belonging to the authenticated user.

Inputs: no named parameters.

list_collectionsA

List all image collections belonging to the authenticated user.

Inputs: no named parameters.

list_library_nichesA

List all available niches in the slideshow library with profile counts.

Inputs: no named parameters.

list_tiktok_accountsA

List TikTok accounts connected via OAuth.

Inputs: no named parameters.

list_tiktok_postsA

List TikTok posts with engagement metrics and aggregated statistics.

Args: timeframe: 7, 30, 90, or "all" (default 30). sort: "recent", "views", "likes", "shares", "comments", "bookmarks". tiktok_account_id: Filter by account. limit: Default 20, max 200. offset: Pagination offset.

Inputs: sort, limit, offset, timeframe, +1 more.

list_videosA

List rendered videos with optional filters.

Args: automation_id: Filter by automation. video_type: "slideshow", "ugc", or "greenscreen". status: "completed", "processing", or "failed". finished: "true" or "false". failed: "true" or "false". created_after: ISO 8601 timestamp. created_before: ISO 8601 timestamp. limit: Max results (default 20, max 100). offset: Pagination offset.

Inputs: limit, failed, offset, status, +5 more.

publish_to_tiktokA

Publish a completed video to TikTok with all settings provided directly (no automation needed).

Note: TikTok limits 6 publishes per 24 hours. Use MEDIA_UPLOAD (draft) to avoid this.

Args: video_id: The video to publish (must be completed). tiktok_account_id: Target TikTok account. upload_type: "slides" (photo post, default) or "video" (.mp4). caption: Post title/caption (max 89 chars for slideshows). description: Description/hashtags (max 4000 chars). post_mode: "DIRECT_POST" (default) or "MEDIA_UPLOAD" (draft). visibility: "PUBLIC_TO_EVERYONE", "SELF_ONLY", "MUTUAL_FOLLOW_FRIENDS", "FOLLOWER_OF_CREATOR". allow_comments: Allow comments (default true). allow_duet: Allow duets (default true). allow_stitch: Allow stitches (default true). auto_music: Let TikTok add trending music (default true, slideshows only). disclose_video_content: Disclose video content. disclose_brand_organic: Disclose brand organic. disclose_branded_content: Disclose branded content. slideshow_image_urls: Override slideshow images with custom URLs.

Inputs: video_id (required), tiktok_account_id (required), caption, post_mode, +11 more.

publish_video_via_automationA

Publish a completed video to TikTok using the automation's saved settings.

The video must be linked to an automation. For standalone publishing, use publish_to_tiktok instead.

Args: video_id: The video ID (must be completed). post_mode: "DIRECT_POST" or "MEDIA_UPLOAD" (TikTok draft). Defaults to automation's setting.

Inputs: video_id (required), post_mode.

run_automationA

Trigger a one-off slideshow generation using the automation's saved settings.

Does not affect the cron schedule. To find the video, note the time before calling, then poll list_videos with automation_id and created_after filters.

Args: automation_id: The automation UUID. hook: Override the hook/topic for this single run. mode: "export" (default, generates video) or "draft_only" (no video).

Inputs: automation_id (required), hook, mode.

search_libraryA

Search real TikTok slideshow profiles for inspiration and competitive research.

At least one filter is required.

Args: q: Text search across slide content. niche: Filter by niche (e.g. "fitness"). product_medium: Filter by product type. region: Account region code (e.g. "US"). audience_region: Audience country code (e.g. "US"). sort: "followers" (default) or "recent". limit: Default 3, max 3. offset: Pagination offset.

Inputs: q, sort, limit, niche, +4 more.

search_pinterestA

Search Pinterest for high-resolution images to use in slideshows.

Paginated up to 5 pages per search query.

Args: q: Search query (e.g. "aesthetic coffee"). cursor: Pagination cursor from a previous response.

Inputs: q (required), cursor.

slideshow_statusA

Check slideshow generation/rendering progress. Poll every 5-10 seconds.

Status flow: draft → generating → rendering → completed (or failed). When a video export exists, also returns video_id and video_status.

Args: slideshow_id: The slideshow ID from generate or create.

Inputs: slideshow_id (required).

update_automationA

Update automation settings, or pause/unpause it.

To pause: set action="pause". To unpause: set action="unpause". For config updates, set the fields you want to change.

Args: automation_id: The automation UUID. action: "pause" or "unpause" (mutually exclusive with other fields). title: Update the title. slideshow_hooks: Replace the hooks list. style: Update the style prompt. language: Update the language. tiktok_account_id: Change target TikTok account. tiktok_post_settings: Update TikTok posting settings. product_id: Change linked product. image_settings: Update image/collection settings.

Inputs: automation_id (required), style, title, action, +6 more.

update_scheduleA

Update one or more schedule jobs on an automation.

For a single update, provide job_id and cron. For batch operations, provide actions array with {type, job_id, cron} objects.

Args: automation_id: The automation UUID. job_id: The job to update (single update mode). cron: New cron expression (single update mode). actions: Array of batch operations [{type: "update"|"delete", job_id, cron}].

Inputs: automation_id (required), cron, job_id, actions.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.7/5.0

    Scored across 26 tools

    Disambiguation5/5

    Every tool has a clearly distinct purpose, and even near-overlapping pairs like generate_slideshow vs. create_slideshow or publish_video_via_automation vs. publish_to_tiktok are explicitly differentiated by AI/manual control and automation vs. direct publication. The descriptions eliminate ambiguity.

    Naming Consistency4/5

    The naming is overwhelmingly consistent: verb_noun snake_case (list_automations, get_video, publish_to_tiktok). The only deviation is slideshow_status, which is a noun_noun format, but all other tools follow the pattern, making this a minor inconsistency.

    Tool Count3/5

    With 26 tools, the server is on the heavier end. The broad domain of automation, slideshow generation, video management, and TikTok publishing justifies many tools, but some redundancy (e.g., delete_schedule duplicates update_schedule's batch delete) makes it feel slightly over-scoped.

    Completeness3/5

    Automation and schedule CRUD are well covered, and the creation-to-publishing pipeline works. However, there are no delete or update operations for videos or slideshows, and collections lack any management tools beyond listing. These gaps prevent full lifecycle management for generated content.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables local-first Instagram reel production entirely over MCP, turning raw footage, URLs, or topics into finished reels with script generation, stock media, voiceover, and captions.
      2
      MIT
    • A
      license
      B
      quality
      C
      maintenance
      Enables any MCP-compatible AI agent to generate videos and images from text or reference frames, chain scenes into seamless multi-scene films with sound and speech, and download finished media with manifests and previews. It also exposes account status, plan limits, history and library management through confirm-gated, safe-by-default operations.
      26
      34
      MIT
    • A
      license
      A
      quality
      C
      maintenance
      Connects any MCP client to a Reelo account so users can create boards, price out AI videos, then generate them and fetch the finished assets, plus plan edits and recreate a video from a link. Most operations are free and only the final run spends account credits, held upfront and refunded if generation fails.
      9
      29 npm
      MIT