{
 "id": "ai-agent",
 "group": null,
 "part_of": null,
 "name": "AI agent",
 "aliases": [
  "agent",
  "software agent",
  "autonomous agent",
  "web agent"
 ],
 "definition_md": "An **AI agent** (or **software agent**) is an autonomous or semi-autonomous program that pursues a user-delegated goal by planning across steps, reading external state, and invoking tools or network endpoints without a human driving each click. On the agentic web, an external agent interacts with a site’s public machine surface across nine lifecycle steps: discovering capabilities ([ARD](/protocols/ard/), [MCP Server Cards](/protocols/mcp-server-cards/)), proving identity ([Web Bot Auth](/protocols/webbotauth/)), checking permissions ([robots.txt](/protocols/ai-crawler-tokens-robots/), [Content Signals](/protocols/cloudflare-content-signals-policy/)), reading structured content ([llms.txt](/protocols/llms-txt/)), invoking actions ([MCP](/protocols/mcp/)), coordinating with peers ([A2A](/protocols/a2a/)), paying ([x402](/protocols/x402/), [AP2](/protocols/ap2/)), presenting UI ([MCP Apps](/protocols/mcp-apps/)), and leaving verifiable audit records.",
 "step": null,
 "related_protocols": [
  "mcp",
  "a2a",
  "ard",
  "webbotauth"
 ],
 "sources": [
  {
   "url": "https://developers.googleblog.com/developers-guide-to-ai-agent-protocols/",
   "date": "2026-03-18",
   "label": "REPORTED"
  },
  {
   "url": "https://aaif.io",
   "date": "2025-12-09",
   "label": "REPORTED"
  }
 ],
 "url": "https://agenticweb.wiki/terms/ai-agent/",
 "okf_url": "https://agenticweb.wiki/okf/terms/ai-agent.md",
 "category": null,
 "type": "term"
}