Mem0 for Hermes: memory tool for Hermes Agent

Mem0 memory plugin for Hermes Agent that routes extraction through Hermes's auxiliary model

What Mem0 for Hermes does

Mem0 for Hermes wraps the Mem0 open-source memory engine, which handles semantic search, deduplication and add, update or delete decisions. Mem0 normally sends its own LLM calls to api.openai.com with an OPENAI_API_KEY. That means a second billing provider on Claude or OpenRouter, and no way to run at all on Codex, where Hermes holds only an OAuth token. This plugin replaces that client so fact extraction goes through agent.auxiliary_client.call_llm, the same path Hermes uses for compression, vision and titles. The calls inherit Hermes credential pools, retries, provider fallbacks and usage accounting.

Embeddings still run outside that route, because Hermes has no embedding path of its own. The default embedder is fastembed, an in-process ONNX model (BAAI/bge-small-en-v1.5, 384 dimensions) that keeps data on the machine and downloads about 67 MB of weights on first use. Ollama, OpenAI and several other embedders can be chosen during hermes memory setup. The vector store and history database stay under $HERMES_HOME/mem0_hermes/, and Mem0 telemetry is off by default.

Key features

Mem0 extraction routed through Hermes's auxiliary model path
Works with Codex OAuth, Anthropic, Bedrock, OpenRouter and local providers
Keyless local default embedder, fastembed
Selectable embedders including Ollama, OpenAI, Hugging Face and Gemini
Local vector store and history database under HERMES_HOME
Interactive setup through hermes memory setup

When to use it

Adding Mem0 memory to a Hermes setup that uses Codex and has no OpenAI API key
Avoiding a second LLM bill for memory extraction on Claude or OpenRouter
Keeping embeddings local with no API key

Who it is for: Hermes Agent users who want Mem0 memory without a separate OpenAI key, especially on Codex or other OAuth-only backends.

How it fits with Hermes Agent

Built for Hermes Agent as a memory provider plugin selected during hermes memory setup. Its LLM calls use the same auxiliary client Hermes uses for its own background tasks.

How to install Mem0 for Hermes

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

hermes plugins install hermes-pro/mem0 --no-enable
hermes memory setup

Requirements: Hermes Agent with a model configured through hermes model; the plugin declares mem0ai and fastembed as dependencies

FAQ

What is Mem0 for Hermes?

Mem0 for Hermes is a memory plugin that brings the Mem0 engine to Hermes Agent. All of Mem0's LLM calls go through the model you already configured in Hermes.

Why use Mem0 for Hermes instead of plain Mem0?

Plain Mem0 sends its LLM calls to the OpenAI API with an OpenAI key, which does not exist on Codex OAuth setups and adds a second provider bill elsewhere. This plugin routes those calls through Hermes, so no extra key is needed.

How do I install Mem0 for Hermes?

Run hermes plugins install hermes-pro/mem0 --no-enable, then hermes memory setup to choose mem0_hermes and an embedder. Start a new session to activate it.