Hermes HQ Subscribe
Memory

Not knowing what your agent remembers: read it, and approve every save

Screenshot: Not knowing what your agent remembers: read it, and approve every save

What it does

Memory is two plain files on your own disk. Open them, fix anything wrong, and flip one setting so nothing gets saved without your approval.

Why it is worth it

Memory is two plain files on your own disk. Reading them and approving each save keeps the agent honest.

Related use cases

See all
MemoryOct 11

Agent memory going stale: the setup that kept it sharp for six months

Only the facts needed in every chat stay in memory; everything else goes to a vault or a skill. A nightly job tidies the memory and skills, and a model-free script takes the backups.

View use case
MemoryOct 14

Cross-chat memory: add the qdrant-memory plugin

The qdrant-memory plugin adds persistent memory to Hermes Agent across separate chats. It automatically extracts and stores facts from your conversations in a local Qdrant database, so Hermes can recall earlier context when you start a new session.

View use case
Screenshot: Smart-home taps buried in apps: a desk gadget that starts the coffee
AutomationOct 9

Smart-home taps buried in apps: a desk gadget that starts the coffee

An M5Stack ESP32 running the Hermes Gadget kit talks to Google Home. Agents hot-swap, so the same screen can run any profile you build.

View use case