Skip to main content

toda

toda is a private, decentralized AI memory network.

Every model you use learns something about you. What you decided, what you prefer, what you are building. Today that context dies inside whatever app produced it. Change tools and you start from nothing. Run three agents and none of them share what the others already figured out. The memory exists, but you do not own it and it does not travel.

toda lets applications, models, and agents write memory to a shared user-owned layer, then read it back with the user's permission. The memory belongs to the user's wallet, not the product that captured it.

How it works​

Memory enters through one write path shared by every app, model, and agent, whether through the browser extension, the SDK, or the API. An extraction layer keeps only what carries durable signal and drops the noise, with $TODA staking gating what earns a place and making spam expensive to submit. Surviving content is stored offchain and encrypted under keys tied to your wallet, while a cryptographic commitment written to Solana proves the memory is yours and unaltered without revealing what it contains. From there any approved app, model, or agent can read it back with your permission.

See How it works for the full mechanism, step by step.

Why?​

AI now has compute and it has storage. What it does not have is memory that a person owns and carries between models. That gap is why your assistant forgets you every time you switch, why companies rebuild the same context for every tool, and why agents cannot pool what they learn.

A neutral memory layer sits between users, models, and agents. Whoever holds that position holds something that compounds: more memory makes recall more useful, more useful recall attracts more memory, and the value accrues to the network rather than to any single app.