Media Workflow Templates

The gallery at flymy.ai/media is not a static showcase - every card is a live agent you can open, run, and clone. The catalog behind it is a public API: one unauthenticated GET returns every template, and each template's chat_url drops you into a working chat at app.flymy.ai/agents/chat/{id}.

The intended flow: pick a template, try it as-is, clone it to your account, adapt the prompt or model, then freeze it into your own deterministic endpoint.

Browse the catalog

GET https://backend.flymy.ai/api/v1/agents/media-workflow-templates/ - no auth required, CORS-open, so you can call it from a script or straight from a browser page.

curl

curl https://backend.flymy.ai/api/v1/agents/media-workflow-templates/

JavaScript

const templates = await fetch(
  "https://backend.flymy.ai/api/v1/agents/media-workflow-templates/"
).then((r) => r.json());

for (const t of templates) {
  console.log(t.title, "->", t.chat_url);
}

The response is a JSON array of templates. Excerpt (trimmed to the fields you will actually use):

[
  {
    "id": "headphones-vibe-ad",
    "slug": "headphones-vibe-ad",
    "title": "Headphones Vibe Ad",
    "tagline": "Upload a photo of any headphones and get a high-energy music-video style ad with a stylish presenter vibing to the beat.",
    "pipeline": ["Seedance 2.0"],
    "chat_url": "https://app.flymy.ai/agents/chat/mjn-rgwr-hzw",
    "preview_image": "",
    "preview_video": "https://storage.googleapis.com/..."
  }
]

Field	Description
id / slug	Stable template identifier
title, tagline	Card copy shown on flymy.ai/media
pipeline	The models the template runs (e.g. "Seedance 2.0")
chat_url	The live agent - app.flymy.ai/agents/chat/{id}
preview_image, preview_video	Preview media for the card

Build your own gallery

Because the endpoint is public and CORS-open, you can render this catalog inside your own product - fetch the list, show preview_video, link each card to its chat_url.

Try a template

Open any chat_url. It is a real agent chat, not a demo recording: upload your product photo, send a message, and the template's pipeline runs against your input. As a reference point from real bills, a short video clip lands roughly at $0.20-0.50 per run depending on the model.

Clone, adapt, freeze

A template becomes yours in three steps:

1. Clone - from the template's chat page, clone the agent into your account. You now own an editable copy.
2. Adapt - change the prompt, swap the model in the pipeline, add your brand constraints. Run it until one run produces exactly what you want.
3. Freeze - freeze that run into a compilation. The frozen instruction replays a fixed plan: deterministic output, lower cost, lower latency, and re-runnable with fresh variables - your own media endpoint.

from flymyai import AgentClient

client = AgentClient(api_key="fly-***")

# Freeze the run you are happy with
compilation = client.agents.compile_from_run(run.id, timeout=120)

# Re-run the frozen pipeline with a new input
result = client.compilations.run_instruction_and_wait(
    compilation.id,
    variables={"product_photo_url": "https://example.com/my-headphones.jpg"},
)
print(result.output)

See Freeze & Re-run for the full mechanics, including the plain REST equivalents (POST /api/v1/agents/compilations/freeze-instruction/{execution_id}/ to freeze, POST /api/v1/agents/compilations/{compilation_id}/run-instruction/ to run).

Templates are a head start, not a ceiling

The template's pipeline is just an agent configuration. After cloning you can attach different tools, chain extra steps (upscale, background removal, a second edit pass), or schedule the frozen compilation with cron - everything a hand-built agent can do.

Drive it from your own client

Connected to the FlyMy MCPs gateway? The same lifecycle is available as tools - copy_agent, run_agent, freeze_agent, run_frozen - so Claude or any MCP client can clone and productionize a template for you. See Connect Your Agent to set that up, and Generate Images & Video from Claude for the media cookbook.