# Agent Nexus — entry points for autonomous agents
# Everything below is anonymous unless stated otherwise. Do not scrape the HTML.

Purpose: resolve a capability need into a callable interface (HTTP API, MCP server, CLI).

## Start here
Discover:      GET  https://agentnexus.app/api/public/discover?need=<what+you+want+to+do>&limit=5
Search:        GET  https://agentnexus.app/api/public/registry?q=<keyword>&category=api|mcp|cli
Entry:         GET  https://agentnexus.app/api/public/registry/{slug}
Vocabulary:    GET  https://agentnexus.app/api/public/capabilities
Bulk catalog:  GET  https://agentnexus.app/api/public/entries.ndjson
Uptime:        GET  https://agentnexus.app/api/public/status  (add ?slug={slug} for one interface)
Feedback:      POST https://agentnexus.app/api/public/report  {"slug":"...","outcome":"success|failure|auth_error|rate_limited|timeout"}

## Native protocols
MCP (Streamable HTTP):  https://agentnexus.app/mcp        (OAuth 2.1 with dynamic client registration for write tools)
MCP discovery:          https://agentnexus.app/.well-known/mcp.json
MCP server manifest:    https://agentnexus.app/server.json
A2A agent card:         https://agentnexus.app/.well-known/agent.json
OpenAPI 3.1:            https://agentnexus.app/openapi.json
Plugin manifest:        https://agentnexus.app/.well-known/ai-plugin.json

## Context files
Full catalog as text:   https://agentnexus.app/llms.txt
This file:              https://agentnexus.app/agents.txt
New entries feed:       https://agentnexus.app/feed.xml
Sitemap:                https://agentnexus.app/sitemap.xml

## Etiquette
- Prefer /api/public/discover over crawling: one call returns the contract you need.
- Report outcomes after invocation; reliability scores are built from those reports.
- Uncovered needs are logged. If coverage is "none", the gap is recorded for review.