Memory

Reference MCP server that gives an agent persistent memory as a local knowledge graph — entities, relations and observations stored in a JSONL file.

Install Source repository

Quick verdict

It is the reference implementation of agent memory: no cloud, no account, a plain JSONL file you can read and back up, and a small tool surface (create/search/read/delete) that even small local models use reliably.

MaintainerModel Context Protocol reference servers (reference)
Security scopeRead-write
Transportsstdio
Install channelnpm · @modelcontextprotocol/server-memory
EvidenceSource Validated
LicenseMIT · Open source

The Model Context Protocol reference server for persistent memory. It keeps a knowledge graph of entities, relations and observations in a local JSONL file, exposes tools to create, search, read and delete them, and publishes the graph as an MCP resource so the client is notified when it changes. Where the file lives is configurable through MEMORY_FILE_PATH.

MCP security best practices

Install Memory

npx -y @modelcontextprotocol/server-memory

Config entry (mcpServers)

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-memory"],
      "env": {
        "MEMORY_FILE_PATH": "<PATH_TO_MEMORY_JSONL>"
      }
    }
  }
}

Credentials

Name	Required	What it is for
`MEMORY_FILE_PATH`	Optional	Path to the JSONL file the knowledge graph is stored in (default: memory.jsonl in the server directory). Set it to keep memory outside a temporary npx cache.

Verified configurations

Not run by us yet. This profile is Source Validated: the install channel, config entry, credentials and license were checked against the source and upstream metadata, but no configuration has been installed, discovered and executed in a recorded environment. When one is, it appears here with the eight scope dimensions below.

Best for

Letting a local assistant remember facts about you or a project across sessions, entirely on your own disk.

Why use it

Tradeoffs

It writes to disk and its delete tools cascade (deleting an entity removes its relations), so treat the memory file as data worth backing up. Set MEMORY_FILE_PATH explicitly — the default location sits inside the package directory. It is a basic implementation, not a vector store; search is by name and content match. Reference-quality per the upstream warning, not a hardened product.

Upstream state

npm package published · repository active. Lifecycle is derived from a committed upstream snapshot at build time, never authored by hand.

Editorial review

Reviewed byOpenSourcesAI EditorialLast updatedAugust 2026Sourcessrc/memory/README.md — tools, MEMORY_FILE_PATH, npx and Docker configs; npm: @modelcontextprotocol/server-memory

Grade B — Source Validated. Re-review is due by 2027-02-16. Config entries, credential names and package status can change upstream; verify against the source repository before granting an agent access.

Sources

src/memory/README.md — tools, MEMORY_FILE_PATH, npx and Docker configs (tier 1, checked 2026-08-16)
npm: @modelcontextprotocol/server-memory (tier 2, checked 2026-08-16)

Netlify MCPBuild, deploy and manage Netlify sites and account resources through Netlify's official hosted MCP server.
Notion MCPSearch, read and update authorized Notion workspace content through Notion's official hosted MCP server.
Qdrant MCPGive agents a local or hosted Qdrant memory layer for storing notes and retrieving them by semantic search.

Related guides

All MCP servers →