nodecode-penny
Penny: your core memory, persona and private notes, read and written by the model. A Nodecode rewrite of HelperLabs/penny-hermes by HelperLabs.
Get it
Tell the agent "add the penny cell", or run nodecode add nc://penny.
Set it up
The cell needs the access token Penny's sign-in hands out, and nothing else. Either export it:
export PENNY_TOKEN=...
or set it in /settings (or the page's Settings tab) under penny: token
for the token itself, token_env for the name of the variable holding it,
or token_file for a file holding it — a Penny CLI's auth.json is read as
it is. A token pasted into the panel is kept in a 0600 file under
~/.nodecode/secrets/; the folder holds no secret.
Everything the model does goes to Penny's own endpoint,
https://app.mypenny.ai/mcp. /penny alone shows the endpoint, whether a
token is set, and the core memory Penny keeps; (help :penny) is the
model's guide to the tools and how to use them.
Use it
Ask for what you want; the model picks the tool:
| Tool | Use |
|---|---|
penny_profile |
the core memory and assistant persona Penny keeps for you |
penny_search |
the private notes nearest a question — past decisions, work, context |
penny_remember |
save one durable fact or decision, with three questions it answers |
Or from the keyboard: /penny how do we deploy runs one search.
Recall is the model's call, not a prefetch: it reads the profile before answering from your preferences or history, and searches when the answer depends on earlier work. Saved notes are yours in Penny — the same notes any other Penny client sees.
What it does not do
The original also injected memory into every turn's prompt automatically and ran its own OAuth login. Here the model asks for memory when it needs it, and the token comes from the section. Penny's scheduled work, spaces sharing and self-update were outside the plugin too.
Credit
MIT, Copyright (c) 2026 Helper Labs. See NOTICE for what was rewritten from the plugin and what was left out; LICENSE is the original's, unchanged.
