--- title: "Peekivo for AI Agents: WebMCP Tools, llms.txt and Markdown" description: "How AI assistants and agents can read Peekivo's public site: llms.txt, a markdown copy of each published page and nine in-page WebMCP tools." canonical_url: https://peekivo.com/en/agents language: en last_updated: 2026-10-02 --- # Using Peekivo from an AI assistant Peekivo's public site has three machine-readable entry points: an llms.txt file at /llms.txt, a markdown copy of every published page at the page's address plus .md, and in-page WebMCP tools in browsers that support them. They let an assistant read what Peekivo is and prepare the waitlist form. Nothing an agent does here starts a watch, sends a message or signs anyone up. ## What exists today - llms.txt at /llms.txt: a short description of Peekivo and links to the markdown copy of each published page, grouped by section. - Markdown copies: each published page has a markdown version at the same address plus .md, for example /en.md for the home page. It is built from the same content as the page and is marked noindex so it does not compete with the page in search. - WebMCP tools: on the public site, pages register tools with document.modelContext (or the older navigator.modelContext) when the browser supports it. A browser without WebMCP shows exactly the same page. ## The WebMCP tools These are all the tools the site registers. Eight only read; one fills in a form field. Their answers come from the same content the pages show, and they return no personal data. - get_product_overview: read-only. What Peekivo is, its canonical description and its current availability. - get_how_it_works: read-only. The steps from a typed sentence to an alert with evidence. - list_watch_examples: read-only, paginated. Example watches with their source, rule and check cadence. The examples are illustrations, not live accounts. - get_faq: read-only. The questions answered on the home page, with their answers. - list_public_pages: read-only, paginated. Public pages with their address, title, whether search engines may index them and their markdown address where one exists. - get_page_markdown: read-only. The markdown copy of a published page, chosen from a fixed list of this site's markdown addresses. - get_current_page: read-only. The published page the person has open: its address, title and, where one exists, its markdown address. - search_site: read-only, paginated. Search the approved public pages by their titles, headings and descriptions. It reads only data already on the page and makes no request. - join_waitlist: fills the email field of the waitlist form on the home page. It never submits: the person must tick the consent box and press the join button. On other pages it returns the home page address. ## The approval rule Agents can read and prepare; people decide. On this site that means join_waitlist only fills in an email address, and the person's own tick and click are the confirmation. In the Peekivo product, which is not available yet, the same rule will apply to watches: a watch will start only after the person approves its plan, the approval will belong to that exact version of the plan, and the server will check it again before anything starts. ## What agents cannot do through these tools - Fetch arbitrary web addresses. get_page_markdown reads only this site's own markdown copies from a fixed list. - Submit a form, join the waitlist on someone's behalf, sign up or sign in. - Reach anyone's watches, alerts or account. Public pages carry no customer data. - Talk to Peekivo's own assistant through page text. In the planned product, the assistant that helps set up a watch is built never to see the text of the pages being watched; page content appears only in cards drawn by Peekivo's own code. ## Planned, not available yet - An API and an MCP server for working with your own watches. They are planned and not available. - A tool that explains a proposed watch plan before you approve it. It is planned and not available. > Peekivo is in early access. Planned items have no release date, and this page will list them only once they exist. ## Crawler policy Public pages are open to search engines and to AI search, answer and training crawlers, including OAI-SearchBot, GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended and CCBot. Every crawler is kept out of the private paths: the app, sign-in, sign-up, invitations, testing pages and the API. The robots.txt file also carries a Content-Signal line that allows search, AI input and AI training. ## Questions about agents and the API Questions about agents and the planned API: hello@peekivo.com. ## Related - [About Peekivo and how it reads pages](https://peekivo.com/en/about) - [Say what you want to be told about](https://peekivo.com/en/ask) - [Proposed plans and limits](https://peekivo.com/en/pricing)