yourmemory

Agentic AI memory MCP server with forgetting-curve decay

TLDR

Run the stdio MCP server (typically launched by an AI client, not a human)

$ yourmemory

Print the executable path and a ready-to-paste MCP config block

$ yourmemory-path

Install the CLI from PyPI

$ pip install yourmemory

Identify the agent that owns a memory namespace via env var

$ YOURMEMORY_USER=[agent_name] yourmemory

yourmemory is a Model Context Protocol (MCP) server that gives AI coding agents long-term memory with biological decay modeled after the Ebbinghaus forgetting curve. Memories that are recalled often gain strength; memories that are not decay over time, mimicking human retention.The server exposes three tools to MCP clients:recall_memory(query)

Hybrid retrieval combining vector similarity, BM25 keyword scoring, and graph expansion, ranked by similarity and strength.

store_memory(content, importance, category?, visibility?)

Persists a new memory with an importance score (0-1). Categories include strategy, fact, assumption, and failure. Visibility may be private or shared.

update_memory(id, new_content, importance)

Re-embeds and replaces an existing memory entry.

It plugs into any MCP-aware client (Claude Code, Cursor, Cline, Windsurf, Continue, Zed) and supports multiple agents sharing one instance with namespace isolation.

PARAMETERS

yourmemory

Start the stdio Model Context Protocol server. Reads JSON-RPC requests on stdin and writes responses on stdout; intended to be launched by an MCP-compatible client.

yourmemory-path

Print the absolute path to the yourmemory executable along with a JSON config snippet ready to paste into an MCP client configuration.

YOURMEMORY_USER (environment)

Identifies the agent owning the memory namespace. Multiple agents on the same instance get isolated private memories plus shared context.

CONFIGURATION

Most clients are configured with a JSON snippet pointing command at the yourmemory binary:

$ {
  "mcpServers": {
    "yourmemory": {
      "command": "yourmemory",
      "env": { "YOURMEMORY_USER": "agent-1" }
    }
  }
}

Run yourmemory-path to get the absolute executable path required by clients that do not resolve PATH at launch time. No external database or service is required; storage is local.

CAVEATS

Not a general-purpose Linux command - yourmemory is meaningful only when launched by an MCP client. Running it directly leaves it waiting for JSON-RPC traffic on stdin. The decay model means memories that are not reinforced are eventually surfaced less; importance and category should be chosen carefully to avoid losing critical facts. Requires Python 3 and pip for installation.

HISTORY

YourMemory was released by sachitrafa in 2025 as an open-source MCP memory server. It cites benchmark results on the LoCoMo long-context conversation benchmark, claiming roughly 16 percentage points better recall than Mem0. The project is written in Python and distributed via PyPI.