The roster · Autumn 2026

Three characters.
One honest match.

TasteCast casts original AI characters for brand briefs. Qloo's taste data decides who fits, every claim shows its evidence, and when the data can't separate two characters, we say so.

Try a demo brief Start your own

One click

Four demo briefs, real Qloo results.

Fictional client brands, each with a real reference brand on Qloo. The verdict shown is the stored result of that run; opening a card replays it with the confirm step and costs no new Qloo calls.

How it works

From a brief to a disclosed endorsement, in three steps.

BriefYour brand, the product, and a well-known reference brand on Qloo with a similar audience. You confirm the exact Qloo entity before anything runs.
RankThe Qloo taste index asks how strongly that brand's fans like each character's music, film, TV and brands, against an everyday-brand baseline. The verdict is "Qloo decided" or "Close call".
PackageScript, shot list, soundtrack and a license quote, each line traced to its Qloo source and labelled #ad · AI-generated character.
For agents

The same loop, as an MCP server.

Any MCP client can run TasteCast as tools: list_demo_briefs, resolve_brand, rank_roster, draft_package and quote_license. The agent has to pass a candidate_id from resolve_brand, so a person (or the calling agent, on their behalf) confirms the exact Qloo entity before anything is ranked.

Streamable HTTP, stateless JSON responses. Same per-network abuse guard as the site (no project-side Qloo cap); license quotes are never charged.

Endpoint https://tastecast.swanky-models.workers.dev/mcp

{
  "mcpServers": {
    "tastecast": {
      "type": "http",
      "url": "https://tastecast.swanky-models.workers.dev/mcp"
    }
  }
}

# or, from a terminal
claude mcp add --transport http tastecast \
  https://tastecast.swanky-models.workers.dev/mcp