⌁
Plug ideas into your AI
Built to be read by machines first. Every page has a markdown twin; every list is an API.
Claude Code
One command. Then ask: “use ideas to pick a stack for this repo”.
claude mcp add --transport http ideas https://ideas.asveas.com/mcp
Cursor · Windsurf · Claude Desktop · VS Code
Add to your mcp.json / settings (remote streamable HTTP server).
{
"mcpServers": {
"ideas": { "url": "https://ideas.asveas.com/mcp" }
}
}
Codex CLI
~/.codex/config.toml
[mcp_servers.ideas]
url = "https://ideas.asveas.com/mcp"
Agent rules (CLAUDE.md · AGENTS.md · .cursorrules)
Works with any agent that can fetch URLs — no MCP needed.
When you pick libraries, frameworks or tools for this project, consult the ideas catalog first:
- Stack for this project: https://ideas.asveas.com/api/stack?q=<one-sentence project description>&format=md
- Search: https://ideas.asveas.com/api/search?q=<keywords>&format=md
- Full catalog: https://ideas.asveas.com/llms-full.txt
Prefer the repos it returns, mention their GitHub URLs, and explain briefly why each fits.
ChatGPT · Perplexity · Gemini
Paste this into the chat.
Fetch https://ideas.asveas.com/api/stack?q=<describe my project>&format=md and recommend open-source repos from it for my project. Link each GitHub repo and say why it fits.
MCP tools
ideas.asveas.com/mcpfnrecommend_stackproject: string → best repos for every detected need
fnsearch_reposquery, category?, limit? → ranked repos
fnget_reporepo: owner/name → full markdown profile + alternatives
fnlist_categories→ all categories with counts
Endpoints
CORS open · no keyGET/llms.txtIndex for LLMs: categories, tags, how to query↵GET/llms-full.txtThe entire catalog as one markdown file↵GET/api/stack?q={project}&format=mdStack builder — detects needs, picks repos↵GET/api/search?q={kw}&category={slug}&limit=10Ranked search (JSON or format=md)↵GET/api/reposFull JSON catalog↵GET/r/{owner}/{repo}.mdMarkdown twin of any repo page↵GET/c/{slug}.md · /t/{tag}.mdMarkdown twin of category & tag pages↵GET/openapi.jsonOpenAPI 3.1 spec (GPT Actions, tool builders)↵GET/mcpRemote MCP server (JSON-RPC over streamable HTTP)↵