Install Mnemosyne
One pip command. Pick the block for your deployment target. Every block is copy-pasteable.
For AI Agents: if you are an LLM installing Mnemosyne for a user, copy the block that matches the deployment target. Each block is self-contained and ends with a verify step.
Standard Python SDK
For any agent or Python application. Includes semantic search (vector + FTS hybrid retrieval).
pip install mnemosyne-memory[embeddings]Hermes Agent plugin
For Hermes Agent. A thin wrapper that adds the plugin manifest; pair it with a core install, then run `hermes config set memory.provider mnemosyne`.
pip install --upgrade 'mnemosyne-memory[embeddings]' mnemosyne-hermesAll features
Embeddings plus optional local LLM and MCP integrations. Use when you want the full surface area.
pip install mnemosyne-memory[all]Minimal
Keyword search only, no embeddings, smallest install. Use when vector search is not needed.
pip install mnemosyne-memoryVerify
Confirm the install worked. Should print the installed version.
python -c "import mnemosyne; print(mnemosyne.__version__)"Try the 4.0 beta
4.0 changes how embedding dimensions are checked and adds a multimodal surface. The beta exists to find out whether the upgrade path works on real databases before it becomes the default.
pip install --pre "mnemosyne-memory[embeddings]"If you are on 3.15.x with an existing database, install the beta and just start it. Tell us whether it came up, and if it refused with a dimension error, whether the documented migration was enough to recover on your own.
Report what happenedThis installs only when you ask for it. A plain install still gives you the current stable release.