Aicontentdrop MCP Server

aicontentdrop

Official CLI and TypeScript SDK for AI Content Drop — generate AI video and images across multiple leading models from the terminal or from code.

No runtime dependencies. Works with npx, no install and no API key for anything read-only.

npx aicontentdrop models --type video --max-credits 15

Install

npm install -g aicontentdrop    # CLI
npm install aicontentdrop       # SDK

Python users can install the equally dependency-free official client from PyPI:

pip install aicontentdrop

Authentication

Reading the model catalogue, quoting costs, and asking questions need no credential. Generating needs an API key you create at Settings → Integrations — any account works, the free tier includes 10 credits.

export ACD_API_KEY=acd_live_…

Each key can carry a daily credit ceiling, so an agent holding it cannot overspend the budget you set.

CLI

acd models --type video --max-credits 20     # catalogue with credit costs
acd cost kling_3_0 --quantity 4              # quote before committing
acd ask "which models generate audio?"       # natural-language lookup
acd me                                       # account and credit balance
acd generate "a golden retriever surfing at sunset" --model kling_3_0 --wait
acd status <video_id>
acd list --limit 20                          # cursor-paginated
acd list --all                               # follows the cursor for you
acd register --name my-agent                 # read-scoped agent token

Every command prints JSON by default so scripts and agents can parse it; add --pretty for humans. Failures print JSON to stderr and exit non-zero (2 failure, 3 rate limited).

The video agent, from a reference

acd analyze ./reference.mp4 --rights own_work     # read a clip or image into words (free, capped)
acd drop "a 15-second vertical ad in this style" \
  --reference ./reference.mp4 --rights own_work \
  --lane auto --confirm-each                      # read it, start the run, print what it saw
acd drop "three product videos for Nimbus Tea"    # a plain run from a written goal
acd drop status <run_id> --notes                  # phase, credits, each piece's score and verdict, retry offers
acd drop retry <run_id> <piece_id> --quote <quote_id> # approve that retry at the quote shown

The SDK maps the same routes: analyzeReference, analyzeReferenceAndWait, getReference, startDrop, getDrop, listDrops, approveDrop, plus the pure helpers retryOffersOf and findRetryOffer that read a run for its retry offers.

Rehearse anything that would cost credits:

acd generate "…" --model kling_3_0 --sandbox   # no provider call, no credits

SDK

import { AiContentDrop } from "aicontentdrop";

const acd = new AiContentDrop({ apiKey: process.env.ACD_API_KEY });

const models = await acd.models({ type: "video", maxCredits: 20 });

// Submit and poll until the render exists. Sends an Idempotency-Key by default,
// so a retry after a dropped response cannot charge twice.
const video = await acd.generateVideoAndWait(
  { prompt: "a golden retriever surfing at sunset", aiModel: "kling_3_0" },
  { onProgress: (v) => console.error(v.status) },
);
console.log(video.video_url);

// Pagination handled for you
for await (const item of acd.allVideos()) {
  console.log(item.id, item.status);
}

Errors throw AcdError with status, code, and retryAfter (on 429).

Billing handoffs

Billing uses the separate scoped billing contract. It is not part of the public product MCP tool or resource surface. New checkout creation is initially available only on a configured billing sandbox deployment; unavailable catalog entries and unfinished subscription/history/management projections return an explicit unavailable response. These commands do not establish production payment readiness.

Create a key in Settings → Integrations and select each permission needed: billing:read for status/history, billing:checkout for checkout links, or billing:manage for portal access. Existing read/generate keys, OAuth tokens, and catalog-only acd_agent_ tokens do not gain billing permission. A generation credit cap does not limit purchases.

acd billing catalog
acd checkout plan --sku subscription:starter:s0:monthly --catalog-version VERSION_FROM_CATALOG --idempotency-key UNIQUE_PURCHASE_INTENT --confirm-purchase
acd checkout topup --sku topup:pack_150 --catalog-version VERSION_FROM_CATALOG --idempotency-key ANOTHER_PURCHASE_INTENT --confirm-purchase
acd billing status CHECKOUT_ID
acd billing status
acd billing history --limit 10
acd billing portal --idempotency-key UNIQUE_PORTAL_INTENT

Checkout requires explicit human initiation and accepted catalog terms. The CLI prints the owned hosted URL and status URL; a human completes payment in the browser. Reuse the exact idempotency key and arguments when retrying one intent. The --sandbox flag simulates generation only and is refused for billing writes; use a configured billing sandbox host with --base-url instead.

The SDK maps these endpoints through billingCatalog, createBillingCheckout, billingCheckout, billingStatus, billingHistory, and createBillingPortal. Supply catalogVersion, skuId, kind, idempotencyKey, and returnTarget to createBillingCheckout. Neither the SDK nor the CLI opens the returned URL.

Install the agent skills

The public repository mirrors fourteen focused skills in its root skills/ directory. Use that directory as the source so discovery selects the product collection directly:

# Preview the product skills without installing anything
npx skills add aicontentdrop/aicontentdrop/skills --list

# Install one capability
npx skills add aicontentdrop/aicontentdrop/skills \
  --skill ai-video-model-picker

# Install the complete AI Content Drop set
npx skills add aicontentdrop/aicontentdrop/skills \
  --skill ai-video-model-picker \
  --skill video-prompt-writer \
  --skill image-generation \
  --skill batch-generation-runner \
  --skill api-error-handling \
  --skill credits-and-plans \
  --skill marketing-studio-ad \
  --skill canvas-board \
  --skill competitor-ad-research \
  --skill video-agent \
  --skill recreate-reference \
  --skill drama-studio \
  --skill short-drama-writer \
  --skill muse-connector

Each skill is independently installable and has a narrow trigger: model selection and quoting, video prompting, image generation, batch execution, error recovery, credits and plans, finished ads, boards, competitor research, the video agent, recreating a reference, short dramas, and hosts that build their own client. See the skills catalogue for the capability map.

For AI agents

• MCP server: POST https://aicontentdrop.com/mcp (Streamable HTTP) — manifest at /.well-known/mcp.json
• Docs-only MCP: POST https://aicontentdrop.com/mcp/docs
• Agent contract: /auth.md
• OpenAPI 3.1: /openapi.json
• Skills: skills/ — model selection, prompt writing, batch runs, error handling, credits, image generation, finished ads, boards, competitor research, the video agent, recreating a reference and short dramas. Also served per skill at https://aicontentdrop.com/skills/<name>/SKILL.md, indexed at /skills/index.json
• Agent plugin: plugin.json + mcp.json (Agent Plugins v1.0.0)
• Coding-agent rules for this repo: AGENTS.md

Links

Documentation https://aicontentdrop.com/developers · TypeScript registry https://www.npmjs.com/package/aicontentdrop · Python registry https://pypi.org/project/aicontentdrop/ · Support support@aicontentdrop.com

MIT licensed.