Memory MCP
Persistent memory MCP server that saves and recalls Claude Code conversation history
- Pricing
- Open source
- Platforms
- CLI, self-hosted
- Runs
- Locally
- Version
- 0.2.1
- Listing updated
- today
Similar MCP servers
About Memory MCP
Memory MCP is an open-source MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.
The README says the service stores Claude Code conversations as episodes, with auto-generated summaries when a session closes. Key knowledge such as decisions, preferences and architecture notes can be pulled out as separate entities and searched later by meaning rather than exact keywords.
Hooks for session start, user prompts, assistant replies and session end handle saving and recall automatically, so relevant past context can be injected into new sessions. Memory is kept in two layers: user-level storage shared across projects and project-level storage isolated per project directory.
It suits developers who work with Claude Code over many sessions and want decisions and context to carry forward. The vector model is roughly 400MB to download on first setup, and the README notes Python 3.10 to 3.13 is required.
Install
Run in a terminal
claude mcp add memory-mcp -- uvx chenxiaofie-memory-mcpAdd to claude_desktop_config.json
{
"mcpServers": {
"memory-mcp": {
"command": "uvx",
"args": [
"chenxiaofie-memory-mcp"
]
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"memory-mcp": {
"command": "uvx",
"args": [
"chenxiaofie-memory-mcp"
]
}
}
}Add to .vscode/mcp.json
{
"servers": {
"memory-mcp": {
"type": "stdio",
"command": "uvx",
"args": [
"chenxiaofie-memory-mcp"
]
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"memory-mcp": {
"command": "uvx",
"args": [
"chenxiaofie-memory-mcp"
]
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"memory-mcp": {
"command": "uvx",
"args": [
"chenxiaofie-memory-mcp"
]
}
}
}Using several servers? Build one config file with the MCP config generator.
Configuration
CLAUDE_PROJECT_ROOT项目根目录路径
What it does
- Episode-based session memory
- Each conversation session is stored as an episode with an automatically generated summary.
- Entity extraction
- Decisions, preferences, concepts and files can be recorded as typed knowledge entities.
- Semantic recall
- Vector-based search retrieves related past episodes and entities for the current task.
- Automatic hooks
- Session and message hooks save context and inject relevant history without manual calls.
- Dual-layer storage
- User-level memory is shared across projects while project-level memory stays isolated.
Good for
- Keep architecture decisions available across weeks of Claude Code sessions
- Recall personal coding preferences without restating them each session
- Resume a feature in a new session with the earlier discussion as context
- Track important project files and their purpose over time
Pricing
Open source
Free and open source under the MIT License.
Memory MCP FAQ
What license is it released under?
The README states the project is under the MIT License.
What Python versions are required?
The README says Python 3.10 to 3.13 is required, as chromadb is not compatible with Python 3.14 or later.
How do I install it for Claude Code?
The README gives a command using uvx to add the server with claude mcp add, plus optional hooks configured in settings.json.
Is Memory MCP free?
Yes. Memory MCP is open source, so you can use and self-host it for free.
How do I add Memory MCP to Claude Code?
Run this command in your terminal: claude mcp add memory-mcp -- uvx chenxiaofie-memory-mcp
What are the best alternatives to Memory MCP?
The closest alternatives to Memory MCP are cogito-mem (Cogi), Hypermnesia MCP, Memory Vault, CogniCore and ContextStream MCP Server.
Where this comes from
- Description: written from the project's README and its MCP Registry entry.
- Install steps: from the official MCP Registry entry.