To use Mnemosyne as Hermes Agent’s external memory provider, install its integration into the Python environment used by the running gateway, select it with hermes config set memory.provider mnemosyne, restart the correct runtime, then test an actual memory write and recall. Docker deployments need special care: put the integration in a persistent side virtual environment, not in a container layer that will be replaced.
Choose the installation path that matches your Hermes deployment
Mnemosyne integrates with Hermes as a plugin implementing Hermes’s MemoryProvider interface. The important first step is to identify the interpreter and home directory used by the running gateway. Installing into whichever Python happens to be active in your terminal may leave Hermes unable to find the package.
- Persistent Docker or image deployment: use a side virtual environment on persistent storage and register the integration in wrapper mode. This keeps it separate from a rebuildable Hermes runtime.
- Local Hermes installation: install and register using the selected Hermes environment and the active user’s home or profile.
- Native Windows: the integration guide recommends persistent wrapper mode. Explicit symlink mode can require Developer Mode or an elevated shell.
For the official Hermes container, the integration guide identifies /opt/data/ as the mounted home rather than ~/.hermes/. Other deployments may use a different mount; substitute the actual persistent mount and home path. Use the same Python major and minor version as the running gateway. These instructions reflect Mnemosyne’s documentation checked on October 4, 2026; Hermes and Mnemosyne setup behavior can change.
Install Mnemosyne in a persistent Docker deployment
The following commands illustrate the documented wrapper-mode route for the official container. Replace /opt/data and the Python executable as needed for your deployment.
#1 Best Overall
export HERMES_HOME=/opt/data
VENV="$HERMES_HOME/.mnemosyne/venv"
python3 -m venv "$VENV"
"$VENV/bin/python" -m pip install --upgrade pip
"$VENV/bin/python" -m pip install 'mnemosyne-memory[embeddings]' mnemosyne-hermes
"$VENV/bin/mnemosyne-hermes" install --mode wrapper --python "$VENV/bin/python"
hermes config set memory.provider mnemosyneThe installer registers the plugin under $HERMES_HOME/plugins. For this wrapper-mode path, Mnemosyne’s installation guide says not to add a separate plugins.enabled entry.
- Restart the actual container or Compose service using its deployment tooling. A gateway-only restart is not a substitute for restarting the deployed service.
- Once the service is running again, run
hermes memory statusin that deployment to check provider state. - Perform the write-and-recall test described below to confirm that memory works end to end.
Install for a local Hermes setup
For a local installation, install the packages into the chosen Hermes Python environment, register Mnemosyne, select it, and restart the gateway:
Rank #2
pip install 'mnemosyne-memory[embeddings]' mnemosyne-hermes
python -m mnemosyne.install
hermes config set memory.provider mnemosyne
hermes gateway restartRun the install and registration steps as the user whose Hermes profile the gateway uses; do not copy a home path from a different account. If the selected virtual environment does not have pip, the integration guide describes using uv pip install --python <hermes-python> ... to target the Hermes interpreter. The package also provides mnemosyne-hermes install; follow the integration guide’s local and operating-system-specific recipe if plugin discovery is unclear.
Choose the right package extras
The base library can be installed without optional extras. The extras change which additional capabilities and dependencies are installed:
| Install choice | What it adds | When to choose it |
|---|---|---|
mnemosyne-memory |
Base library without optional extras. | When you do not need the documented optional profiles. |
mnemosyne-memory[embeddings] |
The standard local semantic-search dependency set. Mnemosyne’s project repository estimates approximately 800 MB for local embeddings; that is a project estimate, not an independent benchmark or guaranteed install size. The repository’s estimate has no publication year stated. | The documented standard choice for local semantic search. |
mnemosyne-memory[all] |
Includes the embeddings profile and adds local-LLM consolidation dependencies, which may bring extra environment or build requirements. | When local-LLM consolidation is needed and the environment can support its dependencies. |
The documentation does not establish a controlled performance, memory-quality, or price comparison among these options. On native Windows, start with [embeddings] unless you specifically need local-LLM consolidation: compatible wheels or a native build toolchain may be needed for the latter’s dependencies.
Activate the provider and handle desktop settings carefully
Select Mnemosyne as the memory provider with:
hermes config set memory.provider mnemosyneMnemosyne’s current integration guide notes a Hermes configuration caveat: declared provider schemas persist non-secret fields in provider-specific JSON or a host store, but cannot currently target memory.mnemosyne through config.yaml. The documented interim options are hermes memory setup or a command such as hermes config set memory.mnemosyne.<key> <value>. This behavior is version-sensitive, so check it against the Hermes distribution you are running.
Rank #4
Verify registration and test a real memory round trip
After restarting the gateway, starting a new session, or restarting the Docker/Compose service as appropriate, check the provider and installed tools:
hermes memory status
hermes tools list
mnemosyne statshermes memory status reports local provider registration and state; it does not prove connectivity or that a memory can be written and recalled. Tool inventory can vary by installed version, so use hermes tools list to see the tools available in your runtime.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →For an end-to-end check, use the installation guide’s functional test: store a disposable test memory, then ask Hermes to recall it. A successful round trip confirms more than provider status alone: Hermes can use the integration to save information and retrieve it again.
What Mnemosyne does during a Hermes session
The integration guide describes three lifecycle hooks:
pre_llm_callinjects relevant working-memory context before a model call.on_session_startinitializes session-scoped state.post_tool_callcaptures tool results when configured.
Mnemosyne’s project documentation calls its integration guide “the canonical Hermes setup guide.” That is the project’s description of its own documentation, not an independent compatibility guarantee.
Disable Mnemosyne or roll back
To disable the external provider while leaving Hermes’s built-in memory active, run:
Free tools Windows power users keep installed
One-click scans. No signup required.
hermes memory offDo not use hermes tools disable memory as a provider rollback: the integration guide says that disables the memory toolset, including provider tools.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
