Skip to main content
Glama
Universal Web Retrieval MCP

Universal Web Retrieval MCP

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault
TAVILY_API_KEYNoTavily keyed mode; absent → official keyless (X-Tavily-Access-Mode: keyless)
ANYSEARCH_API_KEYNoAnySearch keyed mode; absent → official keyless
WEB_RETRIEVAL_TIMEOUTNoOverall chain deadline in seconds (default 45)45
WEB_RETRIEVAL_LOG_LEVELNoLog level; logs go to stderr (stdout is MCP JSON-RPC)WARNING
WEB_RETRIEVAL_MAX_RESULTSNoCap for web_search.limit (default 20)20
WEB_RETRIEVAL_DDGS_TIMEOUTNoDDGS overall timeout in seconds (default 15s)15
WEB_RETRIEVAL_TAVILY_TIMEOUTNoTavily read timeout in seconds (default 20s)20
WEB_RETRIEVAL_ANYSEARCH_TIMEOUTNoAnySearch read timeout in seconds (default 20s)20
WEB_RETRIEVAL_EXTRACT_CHAR_LIMITNoMax fetched-content chars (default 15000)15000
WEB_RETRIEVAL_TAVILY_CONNECT_TIMEOUTNoTavily connect timeout in seconds (default 5s)5
WEB_RETRIEVAL_ANYSEARCH_CONNECT_TIMEOUTNoAnySearch connect timeout in seconds (default 5s)5

Instructions

Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.

This server publishes no instructions, or was last inspected before Glama recorded them.

Capabilities

Features and capabilities supported by this server

Protocol revision2025-11-25

CapabilityDetails
tools
{
  "listChanged": false
}
prompts
{
  "listChanged": false
}
resources
{
  "subscribe": false,
  "listChanged": false
}

Tools

Functions exposed to the LLM to take actions

NameDescription
web_searchA

Web search through a configurable provider chain with automatic failover. Two modes: mode="failover" (default) walks the chain one provider at a time and returns the first success; mode="fanout" queries up to UWR_SEARCH_FANOUT providers concurrently, then merges and rank-fuses the results (provider="fanout", every row carries sources; costs N× provider quota per search). A provider answering OK with zero hits is an empty ANSWER, not an error: failover keeps walking but a query with no matches succeeds with results: [] (attempt {"outcome": "empty"}), never "all providers failed". Default chain: AnySearch -> Tavily -> DDGS; operators can reorder/extend it with UWR_SEARCH_CHAIN (e.g. anysearch,tavily,yuanbao,bocha,deepseek,kimi,ddgs) and pick the default mode with UWR_SEARCH_MODE. Auth is automatic: a configured API key enables keyed mode, otherwise the provider's official keyless access is used. Returns JSON: {success, provider, mode, latency_ms, attempts, results: [{title, url, snippet, published?, site?, sources?}]}. providers overrides the chain order for this one call (e.g. ["bocha"] or ["yuanbao","tavily"]); in fanout mode its first N eligible entries are queried.

web_fetchA

Fetch one URL's page content as Markdown with automatic provider failover (default: AnySearch -> Tavily -> DDGS.extract; operators can reorder the chain with UWR_FETCH_CHAIN). Auth is automatic (keyed when a key is configured, else official keyless). Accepts a single url string, or urls array (max 5) for batch. providers overrides the chain order for this call. Returns JSON array of {url, content, title, content_type, truncated, original_length, provider, mode, latency_ms, attempts} — truncated: true means content was cut at WEB_RETRIEVAL_EXTRACT_CHAR_LIMIT and original_length is the pre-cut character count; failed URLs carry an error field instead. A CDN/CAPTCHA interstitial (Cloudflare "Just a moment…", Akamai "Access Denied") is not returned as content: it is a fallback-eligible error, so the chain tries the next provider. Private, loopback and link-local targets are fetched like any other URL by default; with WEB_RETRIEVAL_BLOCK_PRIVATE=1 they are refused instead. That opt-in check is a target heuristic, not SSRF protection (no protection against DNS rebinding or a redirect hop at the provider). Note: JS-rendered, login-walled or CAPTCHA-protected pages are not guaranteed. PRIVACY: each entry echoes the caller's URL verbatim in url, on success and on failure, so the link stays usable — a credential inside that URL (?token=, ?sig=, ?api_key=) is therefore returned in cleartext, not scrubbed (only an exact value of a provider key configured here is still stripped). error and attempts are scrubbed; url and content are not. The payload enters this conversation and may be recorded in session transcripts or chat logs and sent to the model provider serving it. Prefer short-lived signed URLs and revoke them after the fetch; redaction is not a confidentiality guarantee.

list_providersA

List every provider this server knows: capabilities (search/fetch), whether it is usable right now (configured / key present), whether it is enabled, and its 1-based position in each active chain. Also reports the active search mode (failover/fanout), the fan-out fan width/grace, and the fetch policy (private-network access, block-page detection, extract limit). Never returns key values. Use it to decide what to pass in web_search/web_fetch providers.

Prompts

Interactive templates invoked by user choice

NameDescription

No prompts

Resources

Contextual data attached and managed by the client

NameDescription

No resources

TDQS

A4.4/5.0

Scored across 3 tools

Disambiguation5/5

Each tool targets a distinct action on a distinct resource: search the web, fetch a URL, and introspect the provider configuration. The `providers` parameter explicitly cross-references web_search/web_fetch with list_providers, so the boundary between 'use a provider' and 'inspect providers' is unambiguous.

Naming Consistency5/5

All three names follow a consistent snake_case verb_noun pattern (web_search, web_fetch, list_providers). The prefix/verb split is predictable and unambiguous across the set.

Tool Count5/5

Three tools is a tight, well-scoped surface for a retrieval server: search, fetch, and configuration introspection. There is no filler or redundant tool, and the provider-chain complexity is absorbed into parameters rather than fragmented into extra tools.

Completeness4/5

The core retrieval lifecycle (discover via search, retrieve content via fetch, inspect/adjust providers) is fully covered. Minor gaps remain, such as no caching/session state or structured extraction beyond raw Markdown, but agents can work around these.

Maintenance

ActivityNo data
ResponsivenessOutdated data