Automate AI Media with Kubeez: REST API vs MCP

Creative teams rarely want every asset to start from a manual click. Automation—batch jobs, internal tools, and AI assistants that “run the brief”—needs stable interfaces. Kubeez gives you two: a public REST API for servers and scripts, and a Model Context Protocol (MCP) server for AI clients that speak tools.

This guide explains when to use which, how they fit common workflows, and where to go next in the product and docs.

#Why automation matters for image, video, and audio

Marketing and production pipelines repeat the same shape: pick a model, pass a prompt (and sometimes reference media), wait for a job, then deliver URLs or files downstream. Doing that only through a browser does not scale for CI, back-office tools, or agentic workflows where a model decides the next step.

Kubeez exposes the same generation surface you use in the app—media, music, ads, and account reads—through integrations designed for machines (REST) and for tool-using assistants (MCP).

#A little context: what MCP is

The Model Context Protocol is an open standard for connecting LLM applications to tools, prompts, and resources. The specification describes hosts, clients, and servers and how they coordinate capabilities such as callable tools—useful background when you wire Kubeez into Cursor, ChatGPT-class hosts, or other MCP-compatible clients. See the Model Context Protocol specification for the authoritative overview.

Kubeez’s MCP server is not a replacement for REST; it is the right fit when the caller is an AI assistant that discovers tools and drives jobs through your account.

#Two front doors on Kubeez

#REST API: backends, scripts, and long-lived keys

Use the REST API when you control a server or script and want API key authentication—typical for backends, scheduled jobs, or mobile apps that should not embed an MCP client.

• Base URL: https://api.kubeez.com/v1
• Keys: create and manage keys under API Keys (sk_live_… style secrets).
• Typical media flow: GET /v1/models → optional POST /v1/upload/media for local files → POST /v1/generate/media → poll GET /v1/generate/media/{id} until completed.

Full orientation: REST API overview, detailed headers and limits: REST API, and Swagger UI for interactive reference.

#MCP: assistants, OAuth, and personal access tokens

Use MCP when the caller is an AI assistant or MCP host (for example Cursor or other clients that connect to remote MCP servers).

• Server URL: https://mcp.kubeez.com/mcp
• OAuth (recommended): the client opens a browser; you sign in to Kubeez and approve capabilities—no secret to paste.
• Personal access token: for products that only offer an “API key” field, generate a token under MCP settings and use it as Authorization: Bearer … (same permission model; revocable from active connections).

Concepts and setup: MCP overview and Quick start. Tools follow the same async pattern as the app: start a job, poll status until CDN URLs are ready.

#When to choose REST vs MCP

Prefer REST when you need predictable server-side code, CI, webhooks, or non-assistant integrations; you store an API key in a secret manager and call HTTPS with JSON (and multipart where uploads apply).

Prefer MCP when the user’s workflow is conversational—they ask an assistant to “generate three variants” or “check balance and pick a video model”—and the host already supports MCP tools and OAuth or token fields.

Secrets: API keys are long-lived and belong in vaults; MCP OAuth reduces pasted secrets for interactive clients, while personal access tokens bridge hosts that only accept a static token—rotate and revoke from MCP settings if anything leaks.

#Automation patterns that map cleanly

• Nightly or campaign batches — REST from a worker: loop prompts, respect rate limits, persist job IDs, poll until complete.
• Internal admin or DAM hooks — REST from your backend after an approval step triggers generation.
• Creative copilots — MCP so the assistant calls get_models, generate_media, and get_generation_status in a loop aligned with your instructions.
• Hybrid — prototype with MCP in the IDE, then reimplement the same sequence in REST for production services.

For model choice and capabilities inside either path, the AI models guide remains the best map of what to run for each brief.

#Where to start in the product

• Video and motion: Video generation
• Still images and edits: Media generation
• API keys: API Keys
• MCP connection and PAT: MCP settings

Summary: Kubeez supports programmatic creative work through REST (https://api.kubeez.com/v1) and assistant-driven work through MCP (https://mcp.kubeez.com/mcp). Pick REST for servers and scripts; pick MCP when an AI client should orchestrate tools on your behalf. Read the REST API overview and MCP overview, then ship your first automated job.