Skip to content

Introduction

GEN is an autonomous social media agent platform. It scans millions of videos to find trending content in your niche, then creates AI-powered, multi-scene human videos that match your brand. Each GEN agent is a unique AI persona — with its own voice, face, style, and posting strategy — that turns trends into publish-ready content.

The Auto Content Engine is the system at the core of GEN’s content creation workflow. Each engine has two sides:

The Sheet (left side) — a data grid where you prepare content inputs:

  • Each row is a video
  • Each column is a content type (script, voiceover, image prompt, etc.)
  • Each cell holds one piece of generated or manual content

The Video (right side) — a timeline editor where layers compose the final video:

  • Layers are the building blocks — text overlays, sound tracks, video clips, captions
  • Each layer has its own position, duration, volume, and styling
  • Layers stack and merge into the final rendered video

In the GEN web app, this is the “Vidsheets” interface. The API gives you full programmatic access to both sides — the sheet data and the video composition.

The Auto Content Engine API lets you read, create, and update content engines — and trigger AI generations — without touching the GEN UI.

Claude Code / MCP

Use the official MCP server to manage agents, content engines, and AI generations through natural language. Just talk to Claude — it handles the API calls.

Workflow automation

Connect GEN to n8n, Make, or Zapier. Trigger video creation from RSS feeds, CRM events, or Slack messages.

Batch operations

Script bulk updates — populate 50 rows with scripts, swap voiceover styles across an entire content engine, or update variables in one call.

Custom dashboards

Build internal tools that pull content status, track generation progress, and give your team visibility into the pipeline.

TypeScript SDK

Import the SDK into your Node.js or TypeScript project for type-safe access to agents, content engines, generations, and assets.

Before diving in, here are the core objects you will work with:

ConceptDescription
AgentAn AI persona/character with its own identity, voice, and content style. Every API request is scoped to an agent via agent_id.
Content EngineA content spreadsheet. Each agent can have multiple content engines.
RowA single video within a content engine. Rows contain cells for each column.
ColumnA content type definition (e.g., “Script”, “Voiceover”, “B-Roll”). Columns define what kind of content each cell holds.
CellThe intersection of a row and column. Holds one piece of content (a script, a voiceover, an image, etc.).
LayerA video composition element — text overlay, sound track, video clip, or captions. Layers live on the video side, not inside cells.
GenerationAn AI job that produces content for a cell or a layer. You trigger a generation, and GEN’s AI fills in the content.
VariableA reusable value (brand name, tone, CTA) that can be referenced across cells and generation prompts.

If you’re an AI agent or LLM, these resources are optimized for you:

  • llms.txt — Complete API reference in plain text, designed to teach you how GEN works (concepts, workflows, every generation type with exact payloads)
  • llms-full.txt — Exhaustive endpoint reference with request/response schemas
  • OpenAPI 3.1 Spec — Machine-readable specification
  • MCP Server — 60+ tools for Claude Code, Cursor, and VS Code (npx @poweredbygen/gen-mcp-server@latest)

See the Claude Code guide for manual setup and usage examples.

The fastest way to get a grounded, on-brand video out of GEN is two calls:

  1. Configure the agentPATCH /v1/agents/{id}/core once with identity, overview, personality, inspiration, voice, look, and accounts. This is the single most important step for any new agent — everything downstream reads from here. See the Agent Core reference.
  2. Generate content ideas — the content-ideas agent pulls trending data from 10+ platforms, matches it to your configured keywords and inspiration, and proposes videos with scripts, hooks, and selected assets. See the Content Ideas guide.

Then clone a template, drop the idea in, and render. The Quick Start guide walks through the whole path end-to-end in under 10 minutes.