This server wraps Svmmon's API to generate and manage TikTok slideshows from Claude (or any MCP client). It requires a paid Svmmon subscription.
Check usage (get_usage): View your subscription tier, monthly slideshow limits, reset date, and available TikTok inbox draft slots.
List profiles (list_profiles): See all your Svmmon personas with their ID, name, niche, readiness status, missing requirements, and TikTok link status — only ready: true profiles can generate slideshows.
List presets (list_presets): Discover available slide-structure presets (e.g. listicle) to use during slideshow generation.
Generate hooks (generate_hooks): Produce AI-scored TikTok hook candidates (ranked 0–100 for virality) for a profile, optionally guided by a topic or tone; results are saved to the profile's hook library.
Generate a slideshow (generate_slideshow): Create a complete, rendered TikTok slideshow from a profile and hook; optionally deliver it directly to TikTok as a draft or via Telegram; returns a 24-hour download link.
List recent slideshows (list_slideshows): Browse up to 100 recently generated slideshows with hook text, profile, and creation date.
Fetch a specific slideshow (get_slideshow): Retrieve a slideshow by ID with a fresh 1-hour signed download URL — useful when the original link has expired.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@svmmon-mcp Create a slideshow for my cooking profile"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Generate TikTok slideshows from Claude (and any MCP client) using your Svmmon account.
This is a thin, local stdio MCP server that wraps Svmmon's public /api/v1 REST API. Install it, paste your key, and ask Claude to make slideshows in plain language. It adds no business logic — every tool maps 1:1 onto a real Svmmon endpoint.
Requires an active subscription. The API works on any active paid plan (per-tier volume caps still apply). Install always works; the tools return a clean "requires an active subscription" message until you paste a valid paid key.
app.svmmonapp.com → Settings → API Keys → Svmmon API → Generate key (any paid plan). It looks like svm_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx.
Treat this key like a password. If it's ever exposed, revoke it instantly in Settings → API Keys — the MCP can't revoke it for you.
Claude Desktop — edit claude_desktop_config.json:
{
"mcpServers": {
"svmmon": {
"command": "npx",
"args": ["-y", "svmmon-mcp"],
"env": { "SVMMON_API_KEY": "svm_your_key_here" }
}
}
}Claude Code:
claude mcp add svmmon -e SVMMON_API_KEY=svm_your_key_here -- npx -y svmmon-mcpGenerate 5 hooks for my Marcus profile and make a slideshow from the best one.
Related MCP server: AITuber MCP Server
Tool | What it does |
| Full usage: tier + plan, slideshows used/cap, AI credits, rerolls, shorts/face-swap caps, account caps, per-account TikTok inbox slots. |
| Your profiles with |
| Valid |
| Scored hook candidates for a profile. Spends AI credits, so it is confirm-gated (see below). May return fewer than requested if your AI-credit cap is hit. |
| Generates ONE slideshow (consumes one from your monthly cap — not idempotent, confirm-gated). Returns a 24h download link; optionally delivers to TikTok/Telegram. |
| Studio image generation on your own provider key (BYOK — not idempotent, confirm-gated). |
| Studio video generation on your own provider key (BYOK — not idempotent, confirm-gated). |
| Remake a public TikTok slideshow in one of your profile voices. Costs 2 slideshows + 2 AI credits (confirm-gated). Async: returns a |
| Poll a recreation by |
| Abandon an in-flight recreation (confirm-gated). Not a refund — see below. |
| Your recent slideshows. |
| Fetch a slideshow + a fresh 1h download link by id. |
| Your image collections with per-collection image-fatigue health tiers. |
| One collection's images, with optional 10-min signed URLs so a vision agent can inspect them. |
| Your personal brain for a profile — patterns, voice DNA, next-hook directions (Growth+). |
| The anonymized community brain — what's working across accounts. |
| Post performance — views/likes/shares/comments, day-1 + final. |
| Live TikTok profile + recent-video stats (Growth+). |
| Studio model catalog + which of your provider keys unlock what. |
| Recent studio generations (metadata; provider source URLs may be expired). |
| Your saved studio generations. |
| Your connected TikTok accounts + live connection health (needs_reconnect, connection_error). |
| Your automations — active state, schedule, delivery mode, hook source, hook count, last run status. |
| One automation's full detail by id. |
| Your posting schedule — every slot's time, days, timezone, and which automation fires it. |
| One profile's full detail — bio, audience, tone, CTA, hashtags, readiness, linked accounts/automations. |
| A profile's hook library — text, status, source, used_count, virality_score. |
| Globally trending TikTok posts (not account-specific) — optional hashtag/days filter. |
Every write tool is two-step confirm-gated: the first call returns a preview of the exact consequence and cost and does nothing; you must call again with confirm: true to execute. Surface the preview to the user before confirming.
The four generate_* tools above are confirm-gated the same way, because they spend money — a slideshow slot, AI credits, or your own provider credits. They also need a write-scoped API key, exactly like the tools in this table.
recreate_slideshow is the one tool that does not finish in a single call. It costs 2 slideshows + 2 AI credits (double a normal generate, because it scrapes the source post and runs a vision read on top of writing the deck) and takes 1-4 minutes, so it returns a job_id immediately rather than polling inline — a client-side tool timeout mid-poll would otherwise lose the only handle on a run that may already have spent quota.
Poll get_recreate_status with that job_id every ~10 seconds. Never call recreate_slideshow again to check on a running job; that starts a second paid recreation.
Only one recreation runs at a time per account, shared with the dashboard. A second start returns 429 and spends nothing.
A failure with error.status 422 means the URL was not a slideshow, and that case never billed. Other failures are refunded automatically.
cancel_recreate is not a refund. It only stops the run if the worker has not reached its charge checkpoint. Land it later and the credits are already spent.
Tool | What it does |
| Upload base64 images (max 20) into an owned collection. Counts against your image-storage cap. |
| Add manual hooks (max 50) to a profile's hook library. |
| Permanently delete one hook from a profile's library (irreversible). |
| Edit a profile's safe voice/CTA/hashtag/product fields. Preview shows a current → new diff. |
| Re-post an already-generated slideshow to TikTok/Telegram — no AI call and no monthly quota spent. Uses a TikTok inbox slot. |
| Run an automation now — spends 1 slideshow + AI credits and posts LIVE. Cannot be undone. |
Env var | Required | Default | Notes |
| Yes | — | Your |
| No |
| Override only for self-host/staging. |
"key is missing, invalid, or revoked" (401) — Set SVMMON_API_KEY to a current key from Settings → API Keys.
"requires an active subscription" (403) — Subscribe at app.svmmonapp.com/subscribe. Any active paid plan works.
"a usage cap was reached" (402) — You hit your monthly slideshow or AI-credit cap. Run get_usage.
"Rate limited" (429) — Wait the suggested time if one is shown; otherwise it's a daily AI limit that resets at midnight UTC. The message tells you which.
1.x targets the Svmmon /api/v1 contract. A breaking API change (/api/v2) would ship as 2.x.
This package handles a live, cost-bearing credential. It reads the key from the environment only, never logs it, never puts it in process arguments, and never echoes it in errors. It performs no auto-retries on cost-burning calls. Report security issues to support@svmmonapp.com.
MIT
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
MCP server for Hailuo (MiniMax) AI video generation
MCP server for Kling AI video generation
Hosted MCP server connecting claude.ai, ChatGPT and other AI apps to your own computer
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/coreywessel/svmmon-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server