Generate Tools with AI
Let AI analyze your site and suggest ready-to-use tools automatically.
How it works
Aigentably crawls up to 5 representative pages of your site (homepage + product, cart, checkout, search, account pages discovered via your sitemap or links), extracts structured signals (forms, buttons, frameworks, exposed JS globals), and asks an LLM to propose tool definitions for each page. Suggestions are deduplicated and ranked by importance.
Each suggestion is stamped with the source URL it came from, so you can see which page produced which tool.
If your site already has tools (saved from a previous generation, written by hand, or auto-detected from schema.org and still sitting unsaved), the generator is told about them and is instructed to fill gaps instead of proposing near-duplicates. Any suggestion that exactly matches an existing tool's name is dropped automatically as a safety net.
Using the generator
Open the generate panel
In your site's tool list, click Add Tool → Generate with AI.
Click Generate with AI
The panel shows any previous generation results immediately. Click Generate with AI (or Generate with AI again) to run a fresh generation.
Crawling up to 5 pages plus LLM analysis takes 15-40 seconds. This runs in the background, so it's safe to switch tabs, come back later, or check the site's card on your dashboard, either place shows live progress and picks up where it left off.
Check the crawl report
After generation, an expandable Crawled N pages panel shows each page that was visited, how many tools came from it, and any failures (auth-blocked, non-HTML, timeout).
A cached badge appears if the crawl was reused from cache (see below).
Review the suggestions
Tools are grouped by generation run with a timestamp. Each suggestion shows its name, description, category, and the source page path it was generated from.
- Save: saves the tool directly to your site
- Pencil icon: opens the tool form pre-filled for review before saving
- Save all (N): saves all unsaved tools in the run at once
Detect from schema.org (free, no AI)
Many sites already mark up their pages with schema.org/JSON-LD for SEO (Product, Article, FAQPage, and similar). That markup already contains most of what a WebMCP tool needs, so instead of asking an LLM to re-derive it, Aigentably builds tools straight from it.
This runs automatically the first time you open the generate panel for a site, whether or not you ever click Generate with AI, so you don't need to know this exists to benefit from it. Click Detect from schema.org any time afterward to re-check (useful after you've added new markup to your site).
This is completely deterministic: no LLM call, no AI generation used. It's free for every plan, including Free-tier accounts with no generations left.
Supported types (the same set Google's Rich Results documentation covers):
| Type | Tool created | What it returns |
|---|---|---|
Product | getProductInfo | Name, price, availability, SKU |
Article / BlogPosting / NewsArticle | getArticleInfo | Headline, author, publish date |
FAQPage | getFAQAnswer | Looks up an answer by matching a question |
Recipe | getRecipeInfo | Ingredients, instructions, timing |
Event | getEventInfo | Dates, location, ticket price |
LocalBusiness | getBusinessInfo | Address, phone, hours |
JobPosting | getJobInfo | Title, salary, location |
Each tool's executeJs reads the <script type="application/ld+json"> tags on whatever page it actually runs on, not just the page where the markup was detected, so one getProductInfo tool works across every product page on your site as long as they all carry Product markup.
If a page has no matching schema.org markup, the tool returns { success: false, error: "..." } instead of throwing, so a false positive from a page that stopped carrying the markup fails gracefully for the calling agent.
See Turn Your schema.org Markup Into WebMCP Tools for why this works and where it stops being enough (it only covers read-only lookups, not actions like adding to cart).
Crawl cache
To avoid re-fetching your site on every run, the crawl is cached for 24 hours per site. Repeated Generate with AI again clicks within that window reuse the cached page snapshot but still call the LLM fresh, so you get new tool variants without re-crawling.
If you've changed your site and want to re-crawl immediately, click Force refresh next to the button.
Free vs Pro limits
| Free | Pro | |
|---|---|---|
| Generations | 1 lifetime | 20 per rolling 30 days |
| Tools visible per run | Top 2 | All generated tools (up to 12) |
| History | Recent suggestions | Full history |
| Model | Gemini 2.5 Flash | Gemini 2.5 Pro |
Free users: your first generation is permanent. Locked suggestions are stored and become visible automatically if you upgrade to Pro later, with no re-generation needed.
Pro users who use up their 20 generations before the monthly reset can buy a top-up pack: 50 additional generations for €9, available from the generate panel's error message or from Settings → Subscription. Purchased generations expire when your monthly quota next resets, they don't carry over.
Generation history
Click Generation history → on your site page to see all past suggestions. Each run is listed with a timestamp and how many tools were saved from it.
Crawling pages that aren't publicly linked
The crawler only discovers pages via your sitemap or links from the homepage. A customer dashboard is usually neither, so it never gets picked up automatically, no matter how it's protected.
In the generate panel, expand Crawl pages & authentication (advanced) and list the page paths under Additional pages to crawl, one per line (e.g. /dashboard/orders). These are added to the crawl on top of whatever's auto-discovered.
Password-protected pages
Some pages, like a customer dashboard, also sit behind a login. The crawler can reach these too, if you opt in, in the same section:
Paste a session cookie
Log in with a dedicated test account (not your own admin session) and copy its session cookie into the Session cookie field.
List which paths get it
Add the page paths that need it, one per line, e.g. /dashboard/** or /account/*. The cookie is only ever sent to matching paths, never to the rest of your site.
This closes the authentication gap, but it can't bypass your own bot or firewall protection (Cloudflare, Akamai, and similar). If those still block the crawler, the crawl report shows "blocks automated access" the same as any other blocked page, that's a separate obstacle on your infrastructure's side, not something Aigentably can solve for you. If you want to let us in, allowlist our crawler's user agent, Aigentably-Bot, in your firewall or robots.txt.
Limitations
- JavaScript-only SPAs: falls back to a headless-Chrome render when a plain fetch only sees a JS-hydrated shell, so most client-rendered pages still produce real signals.
- Auth-protected pages: works for paths you've explicitly given a session cookie to (see above). Everything else is crawled as a logged-out visitor.
- Robots.txt: the crawler honors
Disallowrules forUser-agent: *andAigentably-Bot. Block-listed paths are skipped. - Sites blocking automated traffic: pages returning 401/403, or blocked by bot/firewall protection, are reported as "blocks automated access" in the crawl report.
For these cases, use templates or write tools by hand.