Tools
WET (Web Extended Toolkit) – Tools Reference
Section titled “WET (Web Extended Toolkit) – Tools Reference”wet-mcp exposes 5 tools: 3 multiplexed capability tools (search, extract, media), each driven by an action parameter, plus config (server management) and help (full per-tool docs). Results from search, extract, and media are wrapped in an XPIA safety marker – boundary tags plus an instruction to treat the content as data, not commands – since they return externally-fetched web content.
search
Section titled “search”Find information across the web, academic sources, or indexed library documentation. Returns result listings (titles, URLs, snippets) – not full page content. Wrapped as external content.
| Action | Purpose | Key parameters |
|---|---|---|
search |
Web search through SEARCH_BACKENDS (SearXNG, keyed providers, optional-key Firecrawl, or credential-free DuckDuckGo/Startpage) |
query (required), categories, max_results, time_range, include_domains/exclude_domains, expand (LLM query expansion), enrich (fetch real snippets) |
research |
Academic/scientific search (Google Scholar, arXiv, PubMed) | query (required), max_results, time_range |
docs |
Search library documentation, auto-indexing on first query | library (required), query (required), language, version |
docs_resolve |
Free-form library name -> ranked library_id list |
query (library name, required) |
docs_query |
Version-aware docs query honoring a project’s locked library set + token cap | library (required), query (required), version, topic, project_path, limit |
docs_lock_project |
Detect project manifests (pyproject/package.json/go.mod/Cargo.toml) and lock the library set for isolation (Cabinets) | project_path (required) |
similar |
Find pages similar to a URL | query = full URL (required), max_results |
The web-search fallbacks used by research, similar, and docs discovery use
the same configured backend chain. Hosted requests resolve SEARCH_BACKENDS
and matching keys from the current subject’s relay vault; they do not inherit
operator search credentials.
extract
Section titled “extract”Read and return full page content from URLs or local files. Wrapped as external content.
| Action | Purpose | Key parameters |
|---|---|---|
extract |
Get clean content from URLs (markdown/text/html) | urls (required, max 20), format, stealth (anti-bot bypass) |
batch |
Batch extract with per-domain rate limiting | urls (required, max 50) |
crawl |
Deep crawl following links from root URLs | urls (required), depth (default 2, max 5), max_pages (default 20, max 100) |
map |
Discover site URL structure without extracting content | urls (required), depth, max_pages |
convert |
Convert local files (PDF, DOCX, PPTX, XLSX) to Markdown | paths (required) |
extract_structured |
Extract structured data from a page using a JSON Schema + LLM | urls (required), schema (required JSON Schema dict), prompt |
agent |
Multi-step research orchestration: search the web, extract top results, synthesize a cited Markdown answer | query (required), max_urls (default 5, hard cap 20), synthesis_model, token_budget (default 10000) |
interact |
Drive a page with click/fill/submit via a browser automation session | url (required), actions (required list of {type, selector?, description?, value?}), session (persistent session id), screenshot |
diff |
Track content changes across fetches of the same URL(s) | urls (required), refetch (default true – fetch a fresh copy before comparing; set false to compare already-recorded snapshots without a new network fetch) |
BROWSER_BACKENDS selects extraction renderer escalation, including Cloudflare
Browser Run. It does not route interact: that action uses native Patchright
browser sessions and requires their local runtime.
diff requires the web cache enabled (WET_CACHE=true). Returns per-URL change_status (new|same|changed) plus a unified diff (empty when unchanged) and old_fetched_at/new_fetched_at timestamps.
Discover and download media files (images, videos, audio) from web pages. Wrapped as external content.
| Action | Purpose | Key parameters |
|---|---|---|
list |
Scan a page and return media URLs with metadata | url (required), media_type (images|videos|audio|files|all), max_items |
download |
Download media files to local storage | media_urls (required), output_dir (must resolve within the configured download directory) |
Vision/audio/video analysis of downloaded media lives in imagine-mcp’s understand tool, not here.
config
Section titled “config”Server configuration and management, not wrapped as external content.
| Action | Purpose | Key parameters |
|---|---|---|
status |
Show current config, database, embedding/reranker backend, cache, and sync status | – |
set |
Update a runtime setting (log_level, tool_timeout, wet_cache, sync_enabled, sync_folder, sync_interval) |
key, value (required) |
cache_clear |
Clear the web cache | – |
docs_reindex |
Force re-index of a library’s docs | key = library name (required) |
warmup |
Prepare enabled local search, browser, embedding, and rerank resources for the selected runtime | – |
setup_sync |
Configure Google Drive docs sync on eligible non-CF hosts; skipped when DOCS_DB_BACKEND=cf-d1 |
remote_type (default drive) |
setup_status |
Show current credential state and which providers are configured | – |
setup_skip |
Opt into local-only mode (ONNX embed/rerank, no cloud keys) | – |
setup_reset |
Clear all credentials and reset state | – |
setup_complete |
Re-resolve credentials from the environment (re-inits embedding/reranker) | – |
Get full documentation for any tool. Not wrapped as external content.
| Parameter | Values |
|---|---|
tool_name |
search | extract | media | config |
config__open_relay
Section titled “config__open_relay”Opens the HTTP relay/credential setup form in the user’s browser (self-host HTTP mode only; returns stdio_unsupported under stdio). No parameters.