Mnemo
Local-first persistent memory for AI assistants with hybrid search and a knowledge graph
- Pricing
- Open source
- Platforms
- macOS, Windows, Linux
- Runs
- Remotely
- Version
- 2.23.0
- Listing updated
- today
Similar MCP servers
About Mnemo
Mnemo is an open-source MCP server for knowledge and memory. It runs as a hosted remote server, so there is nothing to install locally.
Memories live in one SQLite file under ~/.mnemo/, searched by combining full-text matching and vector search through reciprocal rank fusion, then reranked with a cross-encoder and adjusted for recency and importance. With no API keys it uses local Qwen3 ONNX embedding and reranking models; cloud providers can be configured per task instead, and a configured cloud chain never silently falls back to local.
Capture is typed, with six context types: conversation, fact, preference, skill, task and decision. Similar entries trigger a duplicate warning, and an importance score drives a soft-archive policy with a restore action. Since v2 a bitemporal knowledge graph adds valid-from and valid-to columns, entity resolution and as_of time-travel queries; the README flags a breaking change that default reads now return current state only.
It installs from PyPI as n24q02m-mnemo, ships a separate CLI, and backup is left to rclone.
Install
Run in a terminal
claude mcp add mnemo --env API_KEYS='<API_KEYS>' -- uvx n24q02m-mnemoAdd to claude_desktop_config.json
{
"mcpServers": {
"mnemo": {
"command": "uvx",
"args": [
"n24q02m-mnemo"
],
"env": {
"API_KEYS": "<API_KEYS>"
}
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"mnemo": {
"command": "uvx",
"args": [
"n24q02m-mnemo"
],
"env": {
"API_KEYS": "<API_KEYS>"
}
}
}
}Add to .vscode/mcp.json
{
"servers": {
"mnemo": {
"type": "stdio",
"command": "uvx",
"args": [
"n24q02m-mnemo"
],
"env": {
"API_KEYS": "<API_KEYS>"
}
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"mnemo": {
"command": "uvx",
"args": [
"n24q02m-mnemo"
],
"env": {
"API_KEYS": "<API_KEYS>"
}
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"mnemo": {
"command": "uvx",
"args": [
"n24q02m-mnemo"
],
"env": {
"API_KEYS": "<API_KEYS>"
}
}
}
}Using several servers? Build one config file with the MCP config generator.
Configuration
API_KEYSProvider API keys (format: PROVIDER_API_KEY:key,...). Select models per task with EMBEDDING_MODELS / RERANK_MODELS / LLM_MODELS (CSV provider/model, order = litellm fallback); provider is inferred from the model prefix. Empty embedding/rerank chains select Fastretrieval local ONNX/GGUF models; configured cloud chains never silently fall back to local.
What it does
- Hybrid retrieval
- Full-text and vector search fused, then reranked.
- Typed memory capture
- Six context types with embedding-based duplicate detection.
- Temporal knowledge graph
- Bitemporal entities with history and as_of queries.
- Importance and archiving
- Scores memories and soft-archives stale ones, with restore.
- Local-first storage
- A single SQLite file in WAL mode.
- Optional compression
- LLM summarisation of older memories when a chat model is configured.
- Claude Code plugin
- Recall and commit skills with session hooks.
Good for
- Give a coding assistant memory across sessions
- Store and recall user preferences and decisions
- Track how a fact changed over time
- Consolidate related notes into summaries
Pricing
Open source
Free and open source; cloud model providers are optional
Mnemo FAQ
Does it need API keys?
No. Local ONNX models are used by default; keys are optional for cloud providers.
Where is data stored?
In one SQLite file under ~/.mnemo/.
How do I back it up?
With rclone, since sync is no longer built in.
How do I install it?
pip install n24q02m-mnemo or uvx n24q02m-mnemo.
Is Mnemo free?
Yes. Mnemo is open source, so you can use and self-host it for free.
How do I add Mnemo to Claude Code?
Run this command in your terminal: claude mcp add mnemo --env API_KEYS='<API_KEYS>' -- uvx n24q02m-mnemo
What are the best alternatives to Mnemo?
The closest alternatives to Mnemo are Codebase Memory, XERJ, Claude Code Ultimate Guide, Nautilus Compass Memory and NeuroStack.
Where this comes from
- Description: written from the project's README.
- Install steps: from the official MCP Registry entry.



