All systems normalData updated just nowStatusChangelogSupport
badmintonapi

Use with Claude and agents

An agent skill is a single Markdown file that teaches an AI agent how to use a tool well. Ours covers the whole API: how to get and send a key, which call answers which kind of question, how ids and search work, the limits and how to stay inside them, the errors and what to do about each, and a typical workflow from a plain question to a checked answer. With it, an agent writes correct calls the first time instead of guessing at paths and parameters.

Install it for Claude Code

Two commands
mkdir -p ~/.claude/skills/badmintonapi
curl -s https://badmintonapi.com/skill.md -o ~/.claude/skills/badmintonapi/SKILL.md

Claude Code loads skills from that folder by itself. Ask a badminton question in a project and it reads the skill, then calls the API with your key. Keep the key in BADMINTON_API_KEY in your environment so the agent never has to see it in a file.

Install it in the Claude apps

Download SKILL.md and upload it as a skill in Claude's settings. For live answers inside a chat, connect the MCP server as well (see MCP for AI assistants): the skill teaches the questions, and the server answers them.

Found by agents themselves

Agents that look for skills on their own find ours at /.well-known/agent-skills/index.json. The whole manual is also published as one Markdown file at /documentation/manual.md, a short index at /llms.txt, and the API as OpenAPI 3.1 at /openapi.json, which most agent frameworks can load as a set of tools directly.

Fetch the agent index
curl -s https://badmintonapi.com/.well-known/agent-skills/index.json
Load the API from OpenAPI (Python)
import requests

spec = requests.get("https://badmintonapi.com/openapi.json").json()
for path, ops in spec["paths"].items():
    print(path, ops["get"]["summary"])

Using an agent safely

Every endpoint is read-only, so an agent cannot change anything through the API. Its calls count against your key's limits like any other, so for a long research task give it a per-page size of 200 and let it keep ids it has already found. If an agent will run unattended, use a separate key for it, so you can replace that key without touching your own code.