When web search is enabled, ManyLayers queries the configured search provider before generating a response, fetches the top result pages, and includes a grounding context block in the model’s prompt. Your team gets accurate, up-to-date answers without leaving the workspace.

​
Using web search in chat

If your admin has enabled web search and your team has the permission, you’ll see a Web Search toggle in the chat toolbar. Turn it on before sending a message to ground the response in live search results. After the model responds, citations appear below the answer showing which pages were referenced. You can toggle web search on or off for any individual message — it doesn’t affect the rest of the conversation.

​
Supported providers

Your admin configures which search provider to use. The available options are:
ProviderNotes
SearXNGSelf-hosted; no external API key required
Brave SearchRequires a Brave Search API key
TavilyRequires a Tavily API key
SerperRequires a Serper API key
DuckDuckGoNo API key required; may be rate-limited under heavy use

​
Admin configuration

Configure web search in Admin UI → Settings → Web Search: 
web_search:
  provider: brave          # off | searxng | brave | tavily | serper | duckduckgo
  api_key: "..."           # required for Brave, Tavily, Serper
  base_url: "..."          # required for SearXNG
  top_k: 5                 # maximum search results to fetch per query
When the provider is set to off or cannot be initialized (for example, if the API key is missing), web search grounding is silently skipped and chat continues normally. Users won’t see an error — responses simply won’t include web context.

​
Access control

The web_search feature permission controls which teams can use web search. Users on teams without the permission will not see the web search toggle, even if a provider is configured globally. Admins manage this in Admin UI → Teams → Permissions → Web Search.