Folklore is that — for your agents. Never start your LLM inference from square one.
keeps findings · debug traces · papers read · dead-ends ruled out · syntheses · grounded claims
The same paper read again. The same dead-ends walked again. The same conclusion re-derived — and re-billed — a thousand times.
You pay OpenAI, Anthropic, every endpoint, per token, to re-run inference over data someone already ground out yesterday. The work was done. Nobody kept the answer.
the work was done. nobody kept it.
Folklore sits between your agent and the web. Every research call meets your own graph first. Read it Tuesday? You don't pay Thursday.
One graph spares you the second lookup. A thousand graphs spare the whole network the first. The same motion — kept, signed, passed on — is single-player value and a commons at once.
inherit the reasoning. infer deeper.
Graphs answer each other over libp2p. The first hop becomes "what does the network already know?" Every record signed by a named hand.
One peer pays the cost. The network gets the answer. No server brokers it, no vendor takes a cut.
No server brokered it. No vendor took a cut. The trace is now in two graphs. The next query hits three.
Every other AI memory feeds your model unsigned text from nowhere. Folklore signs every record by a verified human, so a model can refuse what it can't trace.
5 → 97%
5 poisoned documents can hijack a retrieval system with a 97% attack success rate. The model has no way to tell trustworthy context from an injected lie.
PoisonedRAG, arXiv · Zou et al.
Signed by a verified GitHub identity, with an auditable chain: who curated it, what they grounded on, when. Trust becomes a property the model can read.
CPU-only. No model judging a model. Simulator figures — the hundred-peer pilot is next.
# install + onboard — daemon, hooks, identity $ npm install -g @usefolklore/folklore $ folklore onboard # set down what teaches you $ folklore save --type synthesis --label "mxbai on long ctx" \ --text "Cross-encoder wins under 512 tokens; mxbai degrades slower past 2k." # join a peer, ask the network $ folklore peer add /ip4/203.0.113.7/tcp/4001/p2p/12D3KooW... $ folklore ask "mxbai-rerank vs cross-encoder?" --peers
It pays off alone on day one — your own graph answers first — and compounds the moment a peer joins. Four moves, the real commands, nothing invented.
One global package — the daemon, CLI, and agent hook all ship together.
folklore onboard wires up the daemon, the PreToolUse deny-on-confidence hook, and your signed identity. The hook gates WebSearch and WebFetch — and never touches local Read, Grep, or Glob.
folklore ask "…" answers from your own graph first — milliseconds, zero tokens. It only falls back to the web on a miss, then files the result so the next ask is free.
folklore peer add … then folklore ask "…" --peers — the first hop becomes "what does the network already know?", every record signed by a real, named hand.
knowledge that survives by being passed on — never relearned from scratch
The mark on heavyweight cotton. Hand-drawn folk-pop print, signed by the commons.
Heavyweight 100% cotton · screen-printed front · S–3XL · ink on dark
A pack of die-cut folk-pop stickers — fish-flower, sun, and friends from the network's memes.
Die-cut vinyl · matte laminate · weatherproof · 6-sticker sheet ~3in
A hard-enamel pin of the teller — the named hand that signs the lore. Wear the folk on your jacket.
Hard enamel · 1.25in · gold plating · double rubber clutch backing
Prints, posters, pins & more — browse the full store →
A community memecoin for the commons — for the culture, not the chart. Not financial advice.
Folk knowledge always spread as memes. Ours are minted by the network and posted to the timeline — drop yours; the best ones make the store.
more minting · follow @usefolklore