AI Content Drop Developer Documentation

Everything an engineer or an agent needs to generate video and images on AI Content Drop from code.

Start here

Document	What it covers
REST API reference	Every endpoint, request shape, and response shape on `/v1`
Authentication	API keys, agent self-registration, and the discovery documents
MCP server	The Model Context Protocol server, its tools, and its inline UI widgets
GraphQL API	Schema, Relay pagination, typed errors, async jobs, SSE subscriptions, cost and deprecation policy
Sandbox mode	Rehearse a generation without spending credits
Batch operations	Quote or read many things in one request
Rate limits	Limits, headers, and what a 429 tells you
Errors	Every error code across REST, GraphQL, and MCP
Versioning and deprecation	What counts as a breaking change, and how retirement is signalled
Pricing and credits	Plans, credit allocations, and what each model costs

Machine-readable entry points

/openapi.json — OpenAPI 3.1 for the JSON REST API
/openapi-content.json — the unauthenticated content surface
/graphql/schema.graphql — GraphQL SDL; introspection is open at POST /graphql
/.well-known/mcp.json — MCP compatibility discovery manifest
/.well-known/ai-catalog.json — ARD catalogue of every agentic resource we publish
/.well-known/api-catalog — RFC 9727 link set
/llms.txt — curated site index; /developers/llms.txt is the API-scoped version
/skills/index.json — six importable Agent Skills, each served at /skills/<name>/SKILL.md
/plugin.json — Agent Plugins v1 manifest; its MCP servers are at /.well-known/agent-plugin-mcp.json
github.com/aicontentdrop/aicontentdrop — official CLI and SDK source (TypeScript on npm, Python on PyPI), and the agent plugin package (AGENTS.md, plugin.json, mcp.json, skills/)
Append .md to any page URL for its markdown twin, or send Accept: text/markdown

The shape of the product

Describe what you want, pick a model, get a video. 61 video models and 27 image models share one credit balance, at a flat credit cost per generation that is published up front. Credits are charged only when a generation succeeds — safety refusals, provider failures, and timeouts cost nothing, which is why there is no refund endpoint anywhere in this API.

Generation is submit-and-poll. A request returns a job id in about a second; the render takes one to four minutes; you poll until it is done. Nothing in this API blocks on a render.

The five-minute version


# 1. Read the catalogue — no key needed
curl -s https://aicontentdrop.com/v1/models | head

# 2. Quote the cost before committing — still no key
curl -s "https://aicontentdrop.com/v1/models/kling_3_0/cost?quantity=4"

# 3. Rehearse the call for free
curl -sX POST https://aicontentdrop.com/v1/generate/video \
  -H "Content-Type: application/json" -H "X-Sandbox: true" \
  -d '{"prompt":"a cat surfing"}'

# 4. Do it for real
curl -sX POST https://aicontentdrop.com/v1/generate/video \
  -H "Authorization: Bearer acd_live_…" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{"prompt":"a cat surfing","model":"kling_3_0"}'

# 5. Poll the Location header from step 4
curl -s https://aicontentdrop.com/v1/videos/<id> -H "Authorization: Bearer acd_live_…"

Support

Questions go to [email protected] or the contact page.