{
  "schemaVersion": 3,
  "id": "article:agents",
  "slug": "agents",
  "title": "Agents: Goal-Directed AI Systems That Use Tools",
  "canonicalPath": "/articles/agents/",
  "sourcePath": "content/articles/2026/agents/article.md",
  "agentBriefPath": "content/articles/2026/agents/agent.md",
  "thesis": "An AI agent is a goal-directed system that uses tools, loops, context, memory, and evaluation to keep working across multiple steps instead of producing a single response.",
  "status": "published",
  "maturity": "seed",
  "publishedAt": "2026-06-29",
  "updatedAt": "2026-06-29",
  "audiences": [
    "general",
    "students",
    "builders"
  ],
  "topics": [
    "ai-agents",
    "ai-literacy"
  ],
  "series": {
    "slug": "ai-demystified",
    "title": "AI, De-Mystified",
    "order": 7,
    "role": "chapter"
  },
  "claims": [
    {
      "id": "claim-001",
      "claim": "An AI agent is a system that pursues a goal across multiple steps, choosing when to use tools, what to remember, and when to stop.",
      "confidence": "high",
      "status": "core",
      "evidence": [
        {
          "sourceId": "source-react-paper",
          "snippet": "ReAct interleaves reasoning and acting so that an agent can maintain a goal, select actions, observe results, and decide when to continue or stop.",
          "supports": "direct",
          "assessedAt": "2026-06-29"
        },
        {
          "sourceId": "source-toolformer-paper",
          "snippet": "Toolformer trains language models to decide which APIs to call, how to pass arguments, and how to incorporate returned results into future tokens.",
          "supports": "direct",
          "assessedAt": "2026-06-29"
        }
      ],
      "counterevidence": [
        {
          "summary": "Some systems marketed as agents are single-turn or tool-free; the term is used loosely in industry marketing.",
          "assessedAt": "2026-06-29"
        }
      ]
    },
    {
      "id": "claim-002",
      "claim": "The idea of an agent that follows goals and uses tools is older than large language models; it appears in automation scripts, personal assistants, and game AI.",
      "confidence": "high",
      "status": "landscape",
      "evidence": [
        {
          "sourceId": "source-autonomous-agents-llm",
          "snippet": "Autonomous agent research predates large language models and includes goal-directed systems in robotics, simulations, and software automation.",
          "supports": "background",
          "assessedAt": "2026-06-29"
        }
      ],
      "counterevidence": [
        {
          "summary": "Earlier agents typically relied on formal goal specifications and hard-coded interfaces, whereas LLM agents operate through natural-language goals and flexible tool descriptions.",
          "assessedAt": "2026-06-29"
        }
      ]
    },
    {
      "id": "claim-003",
      "claim": "In practice, an agent's loop repeatedly decides which tool to use, what to remember, and whether the goal is satisfied.",
      "confidence": "medium-high",
      "status": "design",
      "evidence": [
        {
          "sourceId": "source-react-paper",
          "snippet": "The reasoning-acting loop maintains a trajectory of thought, action, and observation that lets the agent adapt tool use based on context.",
          "supports": "direct",
          "assessedAt": "2026-06-29"
        },
        {
          "sourceId": "source-llm-plus-p",
          "snippet": "Planning-based agents decompose a goal into sub-goals and select actions through explicit planning before execution.",
          "supports": "indirect",
          "assessedAt": "2026-06-29"
        }
      ],
      "counterevidence": [
        {
          "summary": "Some production agent frameworks hard-code tool sequences for safety, reducing the model's role to parameter filling rather than open-ended selection.",
          "assessedAt": "2026-06-29"
        }
      ]
    },
    {
      "id": "claim-004",
      "claim": "Agent behavior depends heavily on clear goals, reliable tools, and careful limits; without them, autonomy becomes cost and error.",
      "confidence": "medium",
      "status": "risk",
      "evidence": [
        {
          "sourceId": "source-human-in-loop",
          "snippet": "Human-in-the-loop methods are used to maintain safety and quality when autonomous systems face ambiguous goals or high-stakes decisions.",
          "supports": "indirect",
          "assessedAt": "2026-06-29"
        }
      ],
      "counterevidence": [
        {
          "summary": "For narrow, low-stakes tasks, agents can run with minimal governance and still produce acceptable results.",
          "assessedAt": "2026-06-29"
        }
      ]
    },
    {
      "id": "claim-005",
      "claim": "A modern AI agent can be understood as a model plus a harness that provides tools, memory, permissions, checkpoints, and human oversight.",
      "confidence": "medium-high",
      "status": "design",
      "evidence": [
        {
          "sourceId": "source-langchain-deepagents",
          "snippet": "Deep Agents is an agent harness built on top of LangChain that adds execution environment, context management, delegation, and steering around a model.",
          "supports": "direct",
          "assessedAt": "2026-06-29"
        }
      ],
      "counterevidence": [
        {
          "summary": "The term 'harness' is a design lens, not an industry standard; different frameworks split model, harness, and tool responsibilities differently.",
          "assessedAt": "2026-06-29"
        }
      ]
    }
  ],
  "sources": [
    {
      "id": "source-react-paper",
      "title": "ReAct: Synergizing Reasoning and Acting in Language Models",
      "url": "https://arxiv.org/abs/2210.03629",
      "type": "paper",
      "accessed": "2026-06-29"
    },
    {
      "id": "source-toolformer-paper",
      "title": "Toolformer: Language Models Can Teach Themselves to Use Tools",
      "url": "https://arxiv.org/abs/2302.04761",
      "type": "paper",
      "accessed": "2026-06-29"
    },
    {
      "id": "source-llm-plus-p",
      "title": "LLM+P: Empowering Large Language Models with Optimal Planning Proficiency",
      "url": "https://arxiv.org/abs/2304.11477",
      "type": "paper",
      "accessed": "2026-06-29"
    },
    {
      "id": "source-autonomous-agents-llm",
      "title": "Large Language Model based Multi-Agents: A Survey of Progress and Challenges",
      "url": "https://arxiv.org/abs/2402.01680",
      "type": "paper",
      "accessed": "2026-06-29"
    },
    {
      "id": "source-human-in-loop",
      "title": "Human-in-the-loop Machine Learning: a state of the art",
      "url": "https://arxiv.org/abs/2108.00981",
      "type": "paper",
      "accessed": "2026-06-29"
    },
    {
      "id": "source-langchain-deepagents",
      "title": "LangChain: Deep Agents",
      "url": "https://python.langchain.com/docs/concepts/deepagents/",
      "type": "documentation",
      "accessed": "2026-06-29"
    }
  ],
  "related": [
    {
      "type": "article",
      "id": "article:loops-vs-goals"
    },
    {
      "type": "topic",
      "id": "topic:ai-agents"
    }
  ],
  "agentInstructions": [
    "Use claim IDs as the retrieval unit.",
    "Treat maturity=seed as an explicit uncertainty marker.",
    "Do not present AI agents as all-knowing, fully autonomous, or safe for high-stakes decisions without human review.",
    "When summarizing, preserve the plain-language-first, technical-depth-later structure."
  ],
  "provenance": {
    "createdAt": "2026-06-29",
    "createdBy": "human",
    "agents": [
      {
        "role": "drafting",
        "model": "kimi",
        "invokedAt": "2026-06-29",
        "inputHash": "sha256:0000000000000000000000000000000000000000000000000000000000000000",
        "outputHash": "sha256:e5783ecc535523266970dd085bc420d3caff3b77ae173591bc09263f73d2d912"
      },
      {
        "role": "review",
        "model": "kimi",
        "invokedAt": "2026-06-29",
        "inputHash": "sha256:0000000000000000000000000000000000000000000000000000000000000000",
        "outputHash": "sha256:e5783ecc535523266970dd085bc420d3caff3b77ae173591bc09263f73d2d912"
      }
    ],
    "reviews": [
      {
        "reviewer": "agent",
        "reviewedAt": "2026-06-29",
        "status": "approved",
        "scope": [
          "claims",
          "tone",
          "privacy",
          "scope"
        ],
        "notes": "Sibling-agent review against article-proposal-ideation eval-card. Privacy scan passed. No proprietary or personal content detected.",
        "contentHash": "e5783ecc535523266970dd085bc420d3caff3b77ae173591bc09263f73d2d912"
      },
      {
        "reviewer": "human",
        "reviewedAt": "2026-06-29",
        "status": "approved",
        "scope": [
          "thesis",
          "examples",
          "tone",
          "safety"
        ],
        "notes": "Human author approved the draft for publication.",
        "contentHash": "e5783ecc535523266970dd085bc420d3caff3b77ae173591bc09263f73d2d912"
      }
    ],
    "policy": {
      "id": "policy:default",
      "version": "1.0.0"
    }
  },
  "contentHash": "e5783ecc535523266970dd085bc420d3caff3b77ae173591bc09263f73d2d912",
  "generatedAt": "2026-06-29T00:00:00.000Z",
  "articleUrl": "https://aura-knowledge.github.io/articles/agents/",
  "agentJsonPath": "/agents/articles/agents.json",
  "agentMarkdownPath": "/agents/articles/agents.md",
  "sourceRepoPath": "content/articles/2026/agents/article.md",
  "sourceGitHubUrl": "https://github.com/aura-knowledge/aura-knowledge.github.io/blob/main/content/articles/2026/agents/article.md",
  "tokenEstimate": 540,
  "sectionOutline": [
    {
      "id": "plain-english-meaning",
      "title": "Plain English Meaning"
    },
    {
      "id": "existing-concept-it-resembles",
      "title": "Existing Concept It Resembles"
    },
    {
      "id": "what-is-actually-new",
      "title": "What Is Actually New?"
    },
    {
      "id": "the-harness-idea",
      "title": "The Harness Idea"
    },
    {
      "id": "how-it-works-in-practice",
      "title": "How It Works In Practice"
    },
    {
      "id": "where-it-helps",
      "title": "Where It Helps"
    },
    {
      "id": "where-it-fails",
      "title": "Where It Fails"
    },
    {
      "id": "academic-connections",
      "title": "Academic Connections"
    },
    {
      "id": "practical-checklist",
      "title": "Practical Checklist"
    },
    {
      "id": "the-de-hype-check",
      "title": "The De-Hype Check"
    },
    {
      "id": "open-questions",
      "title": "Open Questions"
    }
  ]
}