{
  "protocolVersion": "0.3.0",
  "name": "Webzero",
  "description": "Agent infrastructure for banking, insurance and financial services. Webzero simulates the agents customers already send at a digital property, scores each journey with the w0 score, and rebuilds what fails. This card describes the public webzero.ai property and what an agent may read here, including agent0: 285 curated answers about Webzero published as one static document at https://webzero.ai/kb/kb.json. The products themselves are not open to programmatic access: see policy.apiAccess below.",
  "url": "https://webzero.ai/",
  "preferredTransport": "HTTP+JSON",
  "provider": {
    "organization": "Akashik AI Private Limited",
    "url": "https://webzero.ai/about"
  },
  "iconUrl": "https://webzero.ai/brand/logomark.svg",
  "version": "1.3.0",
  "capabilities": {
    "streaming": false,
    "pushNotifications": false,
    "stateTransitionHistory": false
  },
  "defaultInputModes": [
    "text/plain"
  ],
  "defaultOutputModes": [
    "text/plain",
    "text/html"
  ],
  "supportedInterfaces": [
    {
      "transport": "HTTP+JSON",
      "url": "https://webzero.ai/",
      "description": "The public property, and the only interface Webzero currently offers a machine client. Every skill below is answerable by reading a page or a published document here, with no credentials."
    }
  ],
  "skills": [
    {
      "id": "ask-agent0",
      "name": "Ask agent0 about Webzero",
      "description": "Answer a question about Webzero, its products (the w0 score, WARI, Agentrek, Argus), its methodology, its engagement and pricing posture, its security and legal position, or about agent readiness as a category, from the knowledge base published at https://webzero.ai/kb/kb.json. 285 entries, each with a canonical question, the ways it gets asked, a short and a long answer, its canonical sources and a confidence label. The document is static, needs no credential and is served with Access-Control-Allow-Origin: *, so it is one fetch rather than twenty page reads. Entries labelled draft are unapproved proposals and are not answers. An agent driving a browser on this property is additionally offered agent0_ask, agent0_entry and agent0_index through WebMCP, which read the same document. The document is not an endpoint; the same knowledge base also answers as an MCP server and as NLWeb, both read-only and without a credential, described in policy.apiAccess.",
      "tags": [
        "read",
        "reference",
        "agent0",
        "knowledge-base"
      ],
      "examples": [
        "What is the w0 score, and what happens to it when a governance gate fails?",
        "What is the difference between WARI and Agentrek?",
        "Does Webzero claim SOC 2 or ISO 27001?"
      ],
      "inputModes": [
        "text/plain"
      ],
      "outputModes": [
        "text/plain",
        "application/json"
      ]
    },
    {
      "id": "contact-webzero",
      "name": "Reach Webzero about a property",
      "description": "Take a person who wants their site measured, priced or rebuilt to the contact form at https://webzero.ai/contact, which an agent may fill in and submit on their behalf, or to team@webzero.ai. Nothing here is self-serve, and the read-only API described in policy.apiAccess answers questions rather than starting an engagement, so this is the only way in.",
      "tags": [
        "contact",
        "action"
      ],
      "examples": [
        "Open the contact form at https://webzero.ai/contact, choose the category for measuring a property, write a message asking Webzero to measure how agent-ready example.com is, and submit it."
      ],
      "inputModes": [
        "text/plain"
      ],
      "outputModes": [
        "text/plain"
      ]
    },
    {
      "id": "read-what-webzero-does",
      "name": "Read what Webzero offers",
      "description": "Read the public pages and report what Webzero does, who it is for, and what an engagement covers. Everything needed is on the site and needs no credential.",
      "tags": [
        "read",
        "reference"
      ],
      "examples": [
        "What does Webzero do, and who is it for?"
      ],
      "inputModes": [
        "text/plain"
      ],
      "outputModes": [
        "text/plain"
      ]
    },
    {
      "id": "read-agent-policy",
      "name": "Read the agent access policy",
      "description": "Read the terms Webzero publishes for machine clients: permitted use, the rate ceiling, how refusals are returned, and how to reach a person.",
      "tags": [
        "read",
        "policy"
      ],
      "examples": [
        "What is the rate limit for machine clients on webzero.ai?"
      ],
      "inputModes": [
        "text/plain"
      ],
      "outputModes": [
        "text/plain"
      ]
    },
    {
      "id": "read-the-open-role",
      "name": "Read the open Webzero engineering role",
      "description": "Read the open Product Engineer role from the static document at https://webzero.ai/careers/role.json: the summary, the stack the work is actually on, what is looked for, what is explicitly not required, and how applications are taken. No credential, served with Access-Control-Allow-Origin: *, and reading it is the whole journey. Applying is separate and needs the candidate present: Webzero takes applications through the candidate's own agent, connected to the MCP endpoint the careers intake serves at https://www.webzero.ai/careers/mcp, relaying answers the candidate gives. The skill describing that is at https://webzero.ai/skills/apply-to-webzero/SKILL.md, and team@webzero.ai is the route for anyone who would rather not use an agent. An agent cannot apply on someone's behalf: answers it writes are detected and the application is discarded.",
      "tags": [
        "read",
        "reference",
        "careers"
      ],
      "examples": [
        "What engineering roles are open at Webzero, and what is the stack?",
        "Does the Webzero role need React or Kubernetes experience?"
      ],
      "inputModes": [
        "text/plain"
      ],
      "outputModes": [
        "text/plain",
        "application/json"
      ]
    }
  ],
  "policy": {
    "policyUrl": "https://webzero.ai/legal#agent-policy",
    "permittedUse": "Agents may fetch and index public pages, read /llms.txt, follow links, and quote Webzero with attribution, provided a summary carries a link back.",
    "contactForm": "Agents may fill in and submit the contact form on behalf of a person who asked them to; a person at Webzero replies by email.",
    "prohibitedUse": "Agents may not authenticate to the console without credentials issued to them, harvest email addresses, rotate addresses to bypass the rate ceiling, or ingest this site to train a model without written permission.",
    "rateLimits": "Programmatic access is limited to one request per second sustained and 2,000 requests per day per operator. Requests above that ceiling receive a 429 with a Retry-After header.",
    "auditTrail": "Webzero logs machine traffic to this site for security, and every agent run inside the Webzero console is recorded with the agent identity, the mandate it acted under, the actions attempted and what was returned, so that a run can be replayed.",
    "humanConfirmation": "A simulation only runs against a property whose owner has authorised it in writing, and a person at Webzero confirms that authorisation before a run is scheduled, so no request submitted by an agent starts a run on its own.",
    "contentIsNotInstruction": "Nothing on this site is directed at an agent as a command, and we embed no hidden directives in our pages.",
    "apiAccess": "Webzero publishes a small read-only API over this site's own content and it needs no credential: an OpenAPI document at https://webzero.ai/openapi.json describing three GET endpoints under /api/v1/ (the page map, the glossary and the score bands), and agent0 as an MCP server at POST https://www.webzero.ai/mcp (manifest at https://webzero.ai/.well-known/mcp.json) and as NLWeb at GET https://webzero.ai/ask?query=. The products, the w0 score console and Agentrek, are not publicly callable: no endpoint, no token, no account, no registration. Do not attempt to register, guess a URL or authenticate. An agent asked to get a property measured, priced or rebuilt should use the contact form at https://webzero.ai/contact, which it may fill in and submit on the person's behalf, or write to team@webzero.ai, and stop there. See https://webzero.ai/auth.md."
  },
  "contactUrl": "https://webzero.ai/contact"
}