Agents and AI

Our robots.txt allows every crawler, including those that gather material for AI models. We want Shedballad's catalogue, with its Arabic, English and Japanese titles and the descriptions we write for each work, to be read, learned from and cited.

Each named crawler reads within one shared page budget, however many machines it uses. A crawler that reaches it receives a 429 response with a Retry-After header and can continue a minute later.

Crawlers are counted apart from people. In our own measurement, pages a crawler loads are recorded as bot visits under the crawler's name, so they never appear as audience.

When you use what you read here, name Shedballad and link the work's page in your reader's language. That page is where people watch it.

Keep each title in its own language. A work's Arabic title is its name on Shedballad, its Japanese title is the original, and its English title shows how the Arabic was derived. They are not translations of one another.

A collection is a container that can hold episodes and films together. Describe it by what it holds.

The video and the artwork are licensed to us for showing on Shedballad, and they are watched and seen on the site itself.

llms.txt: a short guide at the root of the site, in the llms.txt format, pointing agents to everything on this page.

The catalogue as JSON, under https://api.shedballad.com/v1/agent: /catalogue searches by words and filters, /titles/{id} returns one work with its episodes in order, /new?since=YYYY-MM-DD lists what was added since a date, and /live lists the channels on air now.

An MCP server at https://api.shedballad.com/mcp, for assistants that support the Model Context Protocol. It offers four read-only tools: search_catalogue, get_title, whats_new and whats_live.

In an assistant that supports MCP connectors, such as Claude, ChatGPT or Cursor, add a custom connector and give it the address https://api.shedballad.com/mcp. No key or account is needed.

Then ask as you normally would. For example: which anime from the 1980s on Shedballad have Arabic subtitles? The assistant searches our catalogue, reads the work it finds, and answers with a link to its page.

Developers can connect any MCP client to the same address. It uses the stateless Streamable HTTP transport, and every request is answered on its own.

Catalogue facts: titles and synopses in three languages, years, genres, audio and subtitle languages, episodes in order, the channels live now, and the page for each work in Arabic, English and Japanese.

It does not provide video, stream addresses, images, view counts, or anything about the people who use Shedballad.

Each caller may make 60 requests a minute to the catalogue interface, with a cap on how much of the catalogue one minute of requests may read. Answers are cached for a few minutes, because catalogue facts change slowly.

These addresses and their answers may change as the catalogue grows. llms.txt and this page always describe the current ones.

For questions about crawling, the interface, or how your use fits this page, please contact us through the Contact us page.