Using an AI agent? Start here
Your agent can look up anime on Anime Meta one show at a time: search by title, then read one show’s scores, airing times and links. It is read-only, needs no key and has no bulk export.
1. Easiest: paste this into your agent
For any agent that can call web addresses (ChatGPT, Claude, Gemini, a script). It tells the agent what to call and how to behave.
You can look up anime on Anime Meta (https://anime-meta.com), one show at a time. - Search: GET https://anime-meta.com/api/v1/search?q=<title> (at most 10 results; each has a slug) - Details: GET https://anime-meta.com/api/v1/anime/<slug> (scores per source, synopsis, airing, links) Rules: read-only JSON, no key, 60 requests per minute, one show per question; never try to download the whole catalogue. When you answer, give the show's page url from the result and say the combined score is Anime Meta's own calculation. The full description is at https://anime-meta.com/openapi.json and https://anime-meta.com/llms.txt.
2. MCP: add the server
For agents that speak the Model Context Protocol (Claude Desktop and Claude Code, Cursor, VS Code, and others). The server is remote, with nothing to install. It offers two tools: search_anime and get_anime.
{
"mcpServers": {
"anime-meta": {
"type": "http",
"url": "https://anime-meta.com/mcp"
}
}
}In Claude Code: claude mcp add --transport http anime-meta https://anime-meta.com/mcp
3. Or call the API yourself
GET https://anime-meta.com/api/v1/search?q=frieren: at most 10 shows.GET https://anime-meta.com/api/v1/anime/<slug>: one show.- Descriptions: /openapi.json (OpenAPI) and /llms.txt.
Fair use
60 requests per minute per address, then HTTP 429 with a Retry-After header. One show per question: please do not walk the catalogue. Cite the page address from each answer. Scores belong to their sites; see About and the terms.