Skip to content
GEN API

The 5-Step Journey

Every GEN API workflow follows the same arc. Whether you’re building a one-off TikTok or automating a 100-video pipeline, the journey has the same five steps. The docs are organized around this arc, and so is the MCP server, the TypeScript SDK, and llms.txt.

Pick a step to jump in, or walk the whole path in order from the Quick Start.

Step 1 — Set Up Your Agent Configure identity, voice, personality, API keys, and workspaces. One PATCH to Agent Core writes everything downstream reads from.
Step 2 — Generate Content Ideas Let the agent propose videos grounded in real trending data from 10+ platforms. Refine, research, set preferences.
Step 3 — Convert Idea to Vidsheet Clone a template that matches the idea's video type. The fastest path to a production-ready engine.
Step 4 — Edit & Generate Fill cells, configure layers, pick creation cards, trigger generations, poll until complete.
Step 5 — Export & Publish Render the final composite, download the MP4, optionally publish to TikTok / Instagram / YouTube.

Mental model

Section titled “Mental model”
Workspace
  └── Agent                     ← Step 1: configure once
        ├── Content Ideas       ← Step 2: data-driven proposals
        └── Auto Content Engine ← Step 3: clone a template
              ├── Sheet         ← Step 4: rows, columns, cells
              │     └── Cells
              │           └── Creation Cards (9 types)
              ├── Video side    ← Step 4: layers per cell
              └── Final Video   ← Step 5: render + publish

Each step produces the input the next step needs. The arc compresses what is otherwise a dozen decisions into a linear path.

Why a journey, not a resource list?

Section titled “Why a journey, not a resource list?”

A resource-oriented API reference (“here are the endpoints for Agents, Rows, Cells…”) is great once you know what you want. But the first time someone touches the API, they don’t ask “what does PUT /v1/autocontentengine/:id/cells/mass_update do?” — they ask “how do I get a video out?”

The 5-step journey is the answer to that question. Every doc surface teaches this arc so you can:

  • Skim the overview for each step and pick the right endpoint
  • Copy-paste a curl chain from llms.txt that walks the full path
  • Let an AI assistant (Claude Code, Cursor) pick the right MCP tool per step

Not sure where to start?

Section titled “Not sure where to start?”