Every use case, hot repo and daily news item on Deduped is also a JSON file. Agents can search the catalog, pull one use case with its steps and agent prompt, and hand it to a coding agent. Fields are only ever added under /api/v1/, never renamed.
Endpoints
| GET | What you get |
|---|---|
/api/v1/index.json | Endpoints and counts |
/api/v1/usecases.json | Every use case, light record (id, repo, title, summary, category, url) |
/api/v1/usecases/<id>.json | One use case: steps, agent_prompt, requirements, risks, caveat. id = repo-slug--use-case-slug |
/api/v1/search.json | Compact search index, one small file for client side matching |
/api/v1/repos.json | Hot AI repos of the last 90 days, best first |
/api/v1/repos/<slug>.json | One hot repo with its use cases |
/api/v1/news.json | Today's AI news brief, each story once |
/api/openapi.json | OpenAPI 3.1 spec |
Example
curl -s https://dedupedai.com/api/v1/search.json | jq '.items[] | select(.t | test("pdf"; "i")) | .id'
curl -s https://dedupedai.com/api/v1/usecases/<id>.json | jq -r .agent_prompt
MCP server
The deduped-mcp MCP server gives Claude, Cursor and any MCP client four tools: search_use_cases, get_use_case, hot_repos and today_news. It only reads the public JSON above over HTTPS. One Python file, no dependencies.
Claude Code
claude mcp add deduped -- uvx deduped-mcp
Claude Desktop or Cursor (mcp.json)
{
"mcpServers": {
"deduped": { "command": "uvx", "args": ["deduped-mcp"] }
}
}
Use and attribution
- Free to read, search and use in your own agent runs and products.
- When you show a use case to a person, link its
urlso they see the full steps, caveats and disclaimer. - The repos belong to their owners and keep their own licenses. Use cases are AI-generated and can be wrong: read the disclaimer.
- Agent guide: llms.txt. Questions: [email protected]
Get 5 new use cases every morning. Free, one short email, unsubscribe anytime.
Get them free