Hermes Supermemory Provider: memory tool for Hermes Agent

Q: Can I self-host Supermemory with it?

Yes. Start Supermemory local with npx supermemory local, add its endpoint as base_url in supermemory.json, then run hermes memory setup and enter the API key the server prints on first boot.

Standalone Supermemory memory provider for Hermes Agent with profile recall and semantic search

What Hermes Supermemory Provider does

Once enabled with hermes memory setup, it prefetches relevant memory before each turn and writes each completed user and assistant turn into one Supermemory document per session per 4-hour window, so the API appends deltas. Failed writes are retried on the next turn, at session end, on /reset or on shutdown, with at-least-once semantics. It exposes four tools: supermemory-save, supermemory-search, supermemory-forget and supermemory-profile, each with a snake_case alias.

Settings live in $HERMES_HOME/supermemory.json and include base_url, container_tag, auto_recall, auto_capture, max_recall_results, profile_frequency, capture_mode and search_mode. A container_tag containing {identity}, such as hermes-{identity}, scopes memories per Hermes profile, and a multi-container mode lets the agent read and write across named containers. It works with the hosted Supermemory API or a self-hosted server started with npx supermemory local. Requests carry an x-sm-source: hermes header that groups memories into a Hermes space in the Supermemory app, which the README calls a routing key, not telemetry.

Key features

Automatic memory prefetch before each turn and capture after each response
Four explicit tools for saving, searching, forgetting and reading the profile
Hybrid, memories-only or documents-only search modes
Profile-scoped containers through the {identity} template
Optional multi-container mode for several named workspaces
Hosted or fully self-hosted Supermemory endpoints

When to use it

Giving a Hermes agent semantic long-term memory with profile recall
Keeping separate memories for each Hermes profile, such as a coder profile
Running memory against a self-hosted Supermemory server for local-only data

Who it is for: Hermes Agent users who want Supermemory as their memory provider, and maintainers who may take over the plugin.

How it fits with Hermes Agent

It is a memory provider for Hermes Agent that was formerly bundled with it, selected through hermes memory setup or the memory.provider setting.

How to install Hermes Supermemory Provider

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

pip install supermemory
hermes memory setup

Requirements: The supermemory Python package and a Supermemory API key, from the hosted service or from a self-hosted Supermemory local server.

Note: The README calls this a handoff repository, a standalone copy published so its upstream maintainers can take it over, and says it is not an official Nous Research plugin.

FAQ

What is Hermes Supermemory Provider?

Hermes Supermemory Provider is a memory provider plugin for Hermes Agent that stores and recalls memories through Supermemory. It supports profile recall, semantic search and explicit memory tools.

Does Hermes Supermemory Provider work with Hermes Agent?

Yes. It is the supermemory provider formerly bundled with Hermes Agent, set up with hermes memory setup or hermes config set memory.provider supermemory. The README says it is not an official Nous Research plugin.

Can I self-host Supermemory with it?

Yes. Start Supermemory local with npx supermemory local, add its endpoint as base_url in supermemory.json, then run hermes memory setup and enter the API key the server prints on first boot.