home / mcp / memorious mcp server
Provides a private, local memory store with semantic recall and per-project isolation.
Configuration
View docs{
"mcpServers": {
"cedricvidal-memorious-mcp": {
"command": "uvx",
"args": [
"memorious-mcp"
]
}
}
}memorious-mcp is a local, private semantic memory MCP server that stores, recalls, and forgets memories entirely on your machine. Built on ChromaDB for vector similarity search, it keeps data private while providing fast, contextually relevant memory retrieval for AI assistants. It supports folder-scoped, per-project storage to keep memories isolated and organized.
You interact with memorious-mcp through an MCP client. Use the three core tools to manage memory: store to save memories with short, embedding-friendly keys; recall to fetch memories based on semantic similarity to a query; and forget to remove memories that match a query. Memories persist across sessions and stay local to your machine.
Prerequisites: ensure you have a runtime capable of running MCP servers and that you can install and run the memorious-mcp server from your environment.
1) Install the MCP runtime you will use to run the server. If you plan to run memorious-mcp via a local setup, you can use the uv or uvx tooling used by MCP clients.
2) Configure the MCP client to include memorious-mcp as a server option. You can use one of the following example configurations.
{
"mcpServers": {
"memorious": {
"command": "uvx",
"args": ["memorious-mcp"]
}
}
}Paths and startup differ by environment. For development or local usage, you may run memorious-mcp directly with your MCP toolchain or integrate it into a per-project workflow. The server uses a local disk-based store under the default storage location and keeps all data on your machine.
Security and privacy: all data stays on your local machine; there are no network requests. Data is stored on disk in the configured storage location (default ./.memorious). Do not store secrets or sensitive credentials in memories unless you accept the local storage risk.
If you encounter issues starting memorious-mcp, verify that the MCP runtime (uvx or uv) is installed and that the command and arguments in your client configuration exactly match the examples above. Ensure the current working directory for local runs points to the memorious-mcp project if required by your setup.
Store facts, preferences, or information with short canonical keys optimized for vector similarity.
Retrieve stored memories using semantic similarity search.
Delete memories matching a query key.