ill13

Analyze Image

Analyze an image using a local vision model. Returns a text description.

Web Search

Search the web using SearXNG. Returns titles, snippets, and URLs.

Fetch Url

Fetch a URL and extract its text content. Falls back to stealth browser (Fortress) for JS-rendered pages, then vision model if DOM extraction fails.

Fetch Rss

Fetch and parse an RSS or Atom feed. Feed URL must be on the configured allowlist.

Fetch Wiki

Search and fetch Wikipedia articles via local Kiwix server. Returns full article text for the top result.

Lookup Sources

Look up saved source URLs for a topic or location. Returns matching URLs from the source registry.

Save Source

Save a source URL under a topic in the source registry for future lookups.

Research

Search the web and auto-fetch results in one call. Accepts a single `query` or multiple `queries` (fanned out in parallel and deduped by URL) for broader coverage in a single call.

Job Search

Search job sites in parallel. Returns raw listings for the LLM to structure. Supports: LinkedIn, Indeed, Dice, BuiltIn, Wellfound, CT State Jobs, USAJobs, JobTensor. For company career pages, use fetch_url directly.

Marketplace Search

Search marketplaces in parallel for items to buy. Returns raw listings for the LLM to structure. Default sites: eBay, Amazon, Craigslist, Facebook Marketplace. Facebook uses GraphQL API (no browser needed).

Community Search

Search community / forum sources for what real people say about a topic. JSON APIs (Hacker News, Stack Exchange, Lemmy, 4chan-archive) return structured posts; the rest (Reddit, KnowYourMeme, Lobsters, Tildes, LessWrong, MetaFilter, Dev.to) are scraped via cookies. Returns raw threads for the LLM to synthesize.

Detect Waf

Navigate a URL with the stealth browser and identify any anti-bot vendor (Cloudflare, DataDome, PerimeterX, Akamai, Kasada, Incapsula). Returns vendor, confidence, recommended strategy, and detection signals. Use this before fetch_protected to plan your approach.

Fetch Protected

Fetch a page that is behind bot protection (Cloudflare, DataDome, PerimeterX, 403, CAPTCHA). Uses the Fortress stealth browser with WAF-aware retry logic. Returns structured status: ok/empty/blocked with WAF vendor identification and strategy. Read the waf field to decide next steps.

Read Page

Get clean reader-mode markdown of any page using the Fortress stealth browser. Handles JS-rendered pages and bot protection. Returns markdown text plus any tables found. Falls back to vision screenshot if DOM extraction fails.

Extract Page

Extract structured data from a webpage: markdown content, tables, metadata (OpenGraph, JSON-LD). Optionally pass a schema object (e.g. {name: null, price: null, sku: null}) to extract specific fields from JSON-LD/OpenGraph. Uses Fortress stealth browser.

Screenshot

Take a screenshot of a webpage using the Fortress stealth browser. Returns the screenshot as a base64-encoded PNG. Optionally saves to a file path. Use full_page=true for full-page captures (useful for long-scrolling pages).

Save Profile

Save the current browser session (cookies + storage state) for a URL as a named profile. Use this after logging into a site, then load_profile later to restore the authenticated session. Profile names are simple strings like 'linkedin' or 'github'.

Load Profile

Load a previously saved browser session profile (cookies + storage). Returns the profile data. After loading, subsequent browser tool calls to the same domain will use the saved cookies automatically (they are merged into the cookies/ directory).

Browser Elements

List interactive elements on a webpage: buttons, links, input fields, and headings. Use this before browser_click or browser_type to see what elements are available. Returns structured element list with visible text and selectors.

Browser Click

Click a button or link on a webpage by visible text or CSS selector. Uses the Fortress stealth browser. Returns the resulting page title and URL after the click.

Browser Type

Type text into an input field on a webpage. Supports field identification by name, placeholder, label, or CSS selector. Optional submit=true to press Enter after typing. Uses human-like typing cadence.

Browser Task

Execute a multi-step browser automation task. Available tasks: login, paginate_collect, infinite_scroll_collect, extract_table, collect_links, download_file, form_fill, search_and_click. Each task accepts specific parameters. Uses Fortress stealth browser.

Crawl Site

Crawl an entire website using the Fortress stealth browser. Follows internal links up to the specified depth and page limit. Handles JS-rendered SPAs. Returns a list of pages with titles and extracted text, plus a sitemap. Use max_pages to control scope (default 20, max 200).

Recon Apis

Reverse-engineer a website's private XHR/Fetch API endpoints. The browser navigates the URL, intercepts all network requests for a specified duration (scrolling to trigger lazy-loaded calls), then returns a list of discovered API endpoints with method, URL, and response type. Secrets are automatically scrubbed. Useful for finding JSON APIs behind JavaScript frontends.