llms.txt Content
# Mnemosyne
> The universal memory layer for any AI agent. Zero-dependency, local-first, fully private. MIT-licensed. SQLite-backed. Works offline.
## Install
For AI agents installing Mnemosyne, use the package that matches the deployment target.
### Standard Python SDK (any agent, includes semantic search)
```bash
pip install mnemosyne-memory[embeddings]
```
### Hermes Agent plugin
`mnemosyne-hermes` is a thin wrapper that adds the plugin manifest and entry
points. It does NOT bundle the core library or embeddings, so install it
alongside one of the packages above, then point Hermes at it:
```bash
pip install --upgrade 'mnemosyne-memory[embeddings]' mnemosyne-hermes
hermes config set memory.provider mnemosyne
```
Verify with `hermes memory status`.
### All features (embeddings + local LLM + MCP)
```bash
pip install mnemosyne-memory[all]
```
### Minimal (keyword search only, no embeddings, smallest install)
```bash
pip install mnemosyne-memory
```
## Verify
```bash
python -c "import mnemosyne; print(mnemosyne.__version__)"
```
Should print the installed version (e.g., `3.10.0`).
## Quick start
```python
from mnemosyne import remember, recall
remember("user-prefers-dark-mode", "The user prefers dark mode for the UI.")
results = recall("user UI preferences", top_k=5)
for r in results:
print(r["content"])
```
## Documentation
- [Install page with all scenarios](https://mnemosyne.site/en/install/)
- [Quick start guide](https://docs.mnemosyne.site/getting-started)
- [Full documentation](https://docs.mnemosyne.site)
- [PyPI: mnemosyne-memory](https://pypi.org/project/mnemosyne-memory/)
- [PyPI: mnemosyne-hermes](https://pypi.org/project/mnemosyne-hermes/)
- [GitHub repository](https://github.com/mnemosyne-oss/mnemosyne)
## Project metadata
- License: MIT
- Python: 3.10+
- Repository: https://github.com/mnemosyne-oss/mnemosyne
- Homepage: https://mnemosyne.site
- Latest stable: mnemosyne-memory 3.15.1