Curated MCP list

The best MCP servers for knowledge and RAG

Knowledge MCP servers give agents a memory. These are the most popular options for RAG and context-rich agent workflows.

What this list covers

Knowledge MCP servers are the context layer of an agentic system. The Memory MCP server exposes a knowledge graph that the agent can read and write. Notion, Confluence, and Obsidian MCPs let the agent search internal docs. Elasticsearch MCPs add semantic search. Together, they turn a stateless model into a teammate that remembers yesterday.

  • Memory MCP for persistent knowledge graphs
  • Notion MCP for docs and databases
  • Confluence MCP for team wikis
  • Obsidian MCP for local-first Markdown notes

The full list

4 Model Context Protocol servers, ordered by adoption.

Memory MCP

Persistent knowledge graph for agents

Knowledge

A knowledge-graph based memory layer that lets agents remember entities, relationships, and observations across sessions. The canonical implementation uses a local JSON store for zero-config persistence.

memorygraphknowledgepersistence
16,700 52,300 78
Install

Notion MCP

Docs, databases, and wikis for agents

Knowledge

Browse, search, and edit Notion pages and databases through MCP. Ideal for agents that need to read product specs, update docs, or summarize team wikis.

notiondocswikidatabase
13,200 44,800 80
Install

Confluence MCP

Pages and spaces for agents

Knowledge

Search, read, and edit Confluence pages and spaces from MCP clients. A practical fit for internal documentation, onboarding guides, and engineering handbooks.

confluencedocswikiatlassian
3,300 9,800 56
Install

Obsidian MCP

Notes and vaults for agents

Knowledge

Read and edit Obsidian vaults through MCP. A natural fit for personal knowledge management agents and research workflows built on local Markdown notes.

obsidiannotesmarkdownvault
2,700 7,800 52
Install

Frequently asked questions

What is the Memory MCP server?

The Memory MCP server is a reference implementation that exposes a knowledge graph of entities, relationships, and observations. Agents can read the graph to recall prior context and write to it to remember new information across sessions.

Can I use MCP for RAG?

Yes. Combine a knowledge MCP (Memory, Notion, Confluence) with a search MCP (Brave, Elasticsearch) and an embedding server. The agent can ground its answers in real, indexed documents.