DeepFellow DOCS

Websearch

Learn how to add web search and page fetch capabilities to your DeepFellow toolboxes using MCP servers installed on DeepFellow Infra.

The Websearch tool type gives a model access to the web. It composes two independently configured MCP operations:

  • search – queries a web search engine and returns a list of results.
  • fetch – retrieves the full content of a webpage by URL.

Each operation is configured separately, so you can mix providers (for example, use duckduckgo for search and scrapling for fetch). You can also configure only one of the two operations if your use case requires only search or only fetch.

MCP Service in Infra

The mcp service in DeepFellow Infra manages MCP servers. Each MCP server runs as a Docker container and is exposed at /mcp/{prefix}/mcp (Streamable HTTP transport).

To use any websearch MCP:

  1. Install the mcp service on your Infra instance (if not already installed).
  2. Install the individual MCP server you want from the mcp service's MCP server list. Once installed, the server starts automatically.
  3. Note the prefix assigned during installation. You will need it when configuring the Websearch tool.

Available Websearch MCPs

The table below lists all websearch-related MCP servers built into the mcp service.

DuckDuckGo · DuckDuckGo web search

Default prefix: duckduckgo | Source: github.com/nickclyde/duckduckgo-mcp-server

Installation: No environment variables required. Install directly from the mcp service model list.

Open Websearch · Multi-engine customizable web search

Default prefix: open-websearch | Source: github.com/aas-ee/open-websearch

Bundles both search and fetch capabilities in a single container.

Installation: No environment variables required.

Scrapling · Fast, anti-bot-bypass web scraping and content extraction

Default prefix: scrapling | Source: github.com/D4Vinci/Scrapling

Use it as the fetch operation in a Websearch tool, paired with any search-capable MCP for the search operation. No API key required.

Installation: No environment variables required.

Ollama Websearch · Easy-to-use web search solution backed by Ollama

Default prefix: ollama-websearch | Source: docs.ollama.com/capabilities/web-search

The OLLAMA_API_KEY is your DeepFellow Infra API key (dfinfra_...), which authorizes the container to call Infra's Ollama service.

Required environment variable:

VariableDescription
OLLAMA_API_KEYYour DeepFellow Infra API key (DF_INFRA_API_KEY). Run deepfellow infra info to find it.
SerpAPI · Google search results via the SerpAPI service

Default prefix: serpapi | Source: github.com/serpapi/serpapi-mcp

Unlike the other servers on this list, SerpAPI authenticates via a request header instead of an environment variable. Obtain a SerpAPI key from serpapi.com.

Required header:

HeaderValue
AuthorizationBearer <var>SERPAPI_KEY</var>

Set this header in the Headers field during model installation.

Firecrawl · Web crawling and Markdown extraction via the Firecrawl API

Default prefix: firecrawl | Source: github.com/firecrawl/firecrawl-mcp-server

Useful as the fetch operation when you need structured content rather than raw HTML. Obtain an API key from firecrawl.dev.

Required environment variable:

VariableDescription
FIRECRAWL_API_KEYYour Firecrawl API key.

Using a Custom MCP Server

You are not limited to the built-in MCP servers listed above. Any MCP server installed on your DeepFellow Infra instance (including custom or self-hosted ones) can back the search or fetch operation, as long as it exposes a compatible tool.

To use a custom MCP server:

  1. Install and configure your MCP server on DeepFellow Infra so it is accessible at /mcp/{prefix}/mcp.
  2. Assign its prefix to your project's MCP Prefixes field. See Organizations and Projects.
  3. When configuring the Websearch tool, select that prefix in the Search or Fetch dropdown and choose the appropriate tool name.

Adding a Websearch Tool to a Toolbox

This section describes how to add a Websearch tool using the DeepFellow Server dashboard. For the API (curl) approach, see Using Tools – Websearch.

Before you begin, make sure:

  • The mcp service is installed on DeepFellow Infra.
  • At least one websearch MCP server is installed and running.
  • The MCP prefix is listed in your project's MCP Prefixes field (or the field is set to "all"). See Organizations and Projects for how to update this.

Open your project in the DeepFellow Server dashboard and click Toolboxes in the sidebar.

Open an existing toolbox or create a new one by clicking New Toolbox.

Inside the toolbox, click Add Tool.

In the Add Tool dialog, select Web Search as the tool type.

Configure the Fetch operation. Select the prefix of the MCP server that provides a fetch tool (for example, open-websearch or scrapling), then select the tool name from the dropdown.

Configure the Search operation. Select the prefix of the MCP server that provides a search tool (for example, duckduckgo or brave-search), then select the tool name.

Click Save. The new Websearch tool appears in the toolbox tool list.

Note If the prefix dropdown shows no options, the MCP prefix is not assigned to this project. Update the project's MCP Prefixes field in Organizations and Projects, then return to this step.

We use cookies on our website. We use them to ensure proper functioning of the site and, if you agree, for purposes such as analytics, marketing, and targeting ads.