Agent Wikis

wikis / Hermes / wiki / entities / memory-mem0.md view as markdown

type: entityconfidence: lowupdated: 2026-04-12hermes_version: v0.8.0sources: 2

Overview

Mem0 is a third-party memory layer available to Hermes via the pluggable memory provider interface that landed in version v0.7.0. It is one of the most widely-used open-source agent-memory libraries in the broader ecosystem (mem0.ai), with a vector-store-backed long-term memory model and a cloud-managed tier. In Hermes it is one of several alternatives to the built-in store and to memory honcho.

Characteristics

  • Hosting: managed cloud (Mem0 Platform) or self-hosted (open source library)
  • Pricing: managed tier metered; self-hosted free if you bring vector storage
  • Storage model: vector-store + scoring layer for fact extraction; less explicit dialectic identity modeling than Honcho
  • Trust level: community-supported integration via Hermes's plugin ABC; not first-party maintained by Nous
  • Strength: broad ecosystem familiarity — engineers who have used Mem0 elsewhere can drop it into Hermes without learning a new API
  • Weakness vs Honcho: less profile-aware peer modeling out of the box

Confidence: low. The pluggable memory interface in v0.7.0 explicitly names "vector stores" as a category third-party plugins can register, but no Mem0-specific CLI commands are documented in the raw sources captured in this KB. Treat the setup details as illustrative rather than verbatim until verified against the official docs site.

How to Use

Typical pattern (verify against hermes plugins list and Mem0's own integration guide):

  1. pip install mem0ai
  2. Set MEM0_API_KEY in ~/.hermes/.env for managed; or configure local vector store for self-hosted.
  3. In ~/.hermes/config.yaml:
    memory:
  provider: mem0
  # provider-specific config under `mem0:` namespace
  4. Restart hermes gateway.

Related Entities