For most of history, markets were closed. Trading happened behind exchange walls with execution opaque to anyone outside; settlement took days; custody was rented from a counterparty. Open access was a marketing line; the architecture was closed by design. Blockchains promised to change that. For simple operations they did. But a general-purpose chain was never built for serious markets: matching priority became a block-producer race, mark prices moved on a cycle separate from execution, liquidations turned into keeper auctions, and order books ran off-chain or too slow to matter. Capital migrated back to custodial venues for speed and certainty. The trade-off was always the same: pick speed and give up self-custody, or pick decentralization and give up venue quality. There was no third option. We are building it. Slowly, and correctly.Trade different.
What Intention is
Intention is a Layer 1 built to run open markets as operations of the protocol. Order matching, margin, liquidation, funding, and settlement are not applications deployed onto a chain, and not an off-chain matching engine reporting results back — they are executed by the network’s own validators, over an ordering that consensus committed, inside a deterministic kernel. The category that names is decentralized clearing and settlement: the functions a market infrastructure performs around a trade — establishing who owes whom, recording it, and making it final — carried out by a protocol rather than by a chain of institutions. The market infrastructure this replaces maps each of those functions onto where it lives here. An exchange is more than that. Front ends, account operations, market listing, support, and commercial terms all sit around it. What the network makes public and verifiable is the part where a discrepancy costs someone money: who filled what, at what price, and who owes whom. The first application running on Intention is Intention Exchange — a perpetual futures venue with a full central limit order book, professional order types, isolated and cross margin, protocol-native liquidations, and on-chain liquidity vaults. It does not implement its own matching engine or risk pipeline; it inherits both by running on the chain. The same substrate is designed to host other order-book markets, including tokenized real-world assets, without changing the execution model.Open by asset, AI-native by protocol
That design removes a specific kind of opacity, and it does so as a consequence of where the trading logic lives rather than as a set of features bolted on top. Because execution runs over an ordering that consensus committed, replaying a block reproduces it exactly. Because the price is certified in the block that consumes it, there is no separate oracle cycle to race. Because liquidation, funding, and insurance are protocol operations rather than contract calls, they run in the same step as the fill that triggered them. Because the kernel emits per-transaction output, every state change can be traced to the transaction that caused it. None of that is a promise the protocol makes separately from its design. It is what a deterministic state machine over a committed ordering produces — which is why the architecture is worth reading as one thing rather than as a feature list. For a human trader, these properties produce a fairer venue: ordering you can audit, prices you can verify, liquidations that follow protocol rules instead of off-chain races, and a per-trade record that exists at the protocol level rather than in someone else’s database. For an agent trading on someone’s behalf, they produce something more basic. Autonomous agents are taking over work humans used to do at the venue layer. Their job is to act on a user’s intention — and they cannot act on a venue they have to model as noise. They need a function: same input, same output, checkable afterwards. A deterministic state machine over a committed ordering is one. AI trading: now and next covers what that is for — what already works when an agent trades on your behalf, what does not yet exist on any venue, and which parts of the protocol are being built for it.Where to go next
Architecture overview
The layers of the network, and how IntentionKernel and IntentionBFT fit together.
AI trading: now and next
What already works when an agent trades on your behalf, and what does not exist yet on any venue.
IntentionKernel
The execution layer: the instruction set, the block pipeline, and why the result is reproducible.
Milestones
The delivery path from testnet to mainnet, and how each stage is verified.
Status
Intention is under active development. Parts of this documentation describe behavior the protocol guarantees by design but that has not yet been exposed through a public mainnet UI. Those pages are marked. The architecture and security sections describe the system as it is being built; the Help Center is largely provisional pending mainnet launch. Developer documentation lives on the developer site and the API reference.No native token currently exists. Documentation does not describe tokenomics, staking incentives, or any token sale. Token-related material will be published only after the asset is live.