Fetch

What Fetch MCP server does

Fetch MCP server is listed in the 🔎 Search & Data Extraction category and is described as handling web content fetching and conversion for efficient LLM usage. That description suggests a use case where an agent retrieves web content and prepares it for language-model processing. However, the supplied MCP tool inventory does not include a web-fetching operation or a URL-specific conversion tool.

Instead, the inventory describes a knowledge graph interface. It includes tools for creating entities, relations, and observations; reading the graph; searching nodes; opening named nodes; and deleting graph data. This mismatch is important when deciding whether the listing fits a web retrieval workflow.

How it works

The listed interface exposes MCP tools that operate on structured graph data. Entities represent items in the graph, relations connect entities, and observations add information to existing entities. Relations are expected to use active voice according to the supplied tool description.

An agent can search for nodes using a query, open selected nodes by name, and read the entire knowledge graph. It can also make batch changes by creating multiple entities or relations, adding observations, or deleting multiple entities and their associated relations. These capabilities describe graph state management rather than direct web navigation or page extraction.

Tools and capabilities

The supplied tool list contains nine operations:

• create_entities: Create multiple entities in the knowledge graph.
• create_relations: Create multiple relationships between entities.
• add_observations: Add observations to existing entities.
• delete_entities: Delete entities and their associated relations.
• delete_observations: Remove specific observations from entities.
• delete_relations: Delete multiple graph relations.
• read_graph: Read the complete knowledge graph.
• search_nodes: Find graph nodes based on a query.
• open_nodes: Open specific nodes by name.

For a developer evaluating a Fetch MCP server, the key consideration is the difference between the advertised web-content purpose and these graph-oriented tools. The listed operations could support an agent that maintains or queries structured knowledge, but the provided material does not establish that they fetch web pages, follow links, extract page text, or convert HTML.

Limitations and notes

The repository excerpt identifies its servers as reference implementations intended to demonstrate MCP features and SDK usage. It also warns that they are educational examples rather than production-ready solutions, and that developers should evaluate security requirements and add safeguards for their own threat models.

The supplied material does not specify a Fetch-specific installation command, required environment variables, authentication method, supported URL types, output format, request limits, or content-conversion behavior. It also does not identify client compatibility for this particular server. The repository gives general examples for configuring MCP servers with Claude Desktop, but those examples are for other servers and should not be treated as Fetch configuration.

Before adopting Fetch MCP server for web retrieval, inspect the implementation and confirm its actual registered tools. If the knowledge graph tool list is authoritative, this entry belongs to a graph-memory workflow despite its web-fetching description.