All pages

Start here

SuperGen is an MCP server. Your agent calls the tools, SuperGen renders the image or video and publishes it to your connected accounts. There is nothing to install — you point your agent at a URL and give it your API key.

Connect your agent

Create a key in Settings → API. The same key authenticates both the MCP server and the REST API, and it is shown once — regenerate if you lose it.

Claude Code

One command. Run it in any project; add --scope user to make it global.

claude mcp add --transport http supergen https://mcp.supergen.tech/mcp \
  --header "Authorization: Bearer sk_your_key"

Claude Desktop

Settings → Connectors → Add custom connector. Paste the URL, then add the header.

URL:    https://mcp.supergen.tech/mcp
Header: Authorization: Bearer sk_your_key

Cursor

Add it to ~/.cursor/mcp.json (global) or .cursor/mcp.json in the project.

{
  "mcpServers": {
    "supergen": {
      "url": "https://mcp.supergen.tech/mcp",
      "headers": { "Authorization": "Bearer sk_your_key" }
    }
  }
}

ChatGPT

Settings → Connectors → Create. Remote MCP servers need a paid plan with developer mode enabled.

URL:  https://mcp.supergen.tech/mcp
Auth: Bearer sk_your_key

Anything else that speaks MCP

Streamable HTTP transport, bearer auth. No SDK of ours to install.

POST https://mcp.supergen.tech/mcp
Authorization: Bearer sk_your_key

Your first post

Ask in plain language — “make a 9:16 video of a paper boat in a stream after rain and post it to my channel”. This is what your agent does with that:

  1. 1

    Connect an account to publish to

    connect_account returns a link you open once to authorise the platform. Or connect it from Accounts.

  2. 2

    Ask for a render

    generate_video(
      prompt="paper boat in a stream after rain",
      aspect_ratio="9:16",
    )
    → {"job_id": "job_ab12", "status": "queued"}
  3. 3

    Poll until it is done — polling is free

    get_job("job_ab12")
    → {"status": "succeeded", "assets": [{"url": "https://…/render.mp4"}]}

    Credits are spent once, when a render succeeds. A failed render bills nothing, and retrying with the same idempotency_key returns the first job instead of rendering twice.

  4. 4

    Publish it, now or later

    create_post(
      account_id="acct_x",
      media_urls=["https://…/render.mp4"],
      scheduled_time="2026-08-01T09:00Z",   # omit to post immediately
    )

    Scheduling runs on our side, so your agent does not have to stay awake for it. Posts we generated are declared as AI-made where the platform requires it (TikTok, YouTube) without you passing anything.

What it costs

Generation spends credits; reading, polling, scheduling and publishing do not. Every model’s price is listed on Models & prices, and estimate_cost answers for a specific call before you make it. New accounts get 25 free credits, no card.

Everything the agent can call

All 41 tools — generation, publishing, scheduling, analytics, comments and DMs — with every parameter, is on the Tool reference.

Next
Credits & billing
What spends credits, what never does, and how a price is worked out.