> ## Documentation Index > Fetch the complete documentation index at: https://docs.inkwell.finance/llms.txt > Use this file to discover all available pages before exploring further. # Architecture > How Dagon's matching engine operates on ciphertexts it cannot decrypt. > **CONFIDENTIAL & PROPRIETARY © 2026 Inkwell Finance, Inc. All Rights Reserved.** > > This document is for informational purposes only and does not constitute legal, tax, or investment advice, nor an offer to sell or a solicitation to buy any security or other financial instrument. Any examples, structures, or flows described here are design intent only and may change. Dagon's matching happens entirely on ciphertext. Neither the operator nor anyone observing the chain sees the plaintext inputs, the clearing search, the fills, or the balances. The FHE evaluator operates blind. This section covers two things at the conceptual level: How a batch of participants fits into a compact ciphertext layout, and why that layout lets the matching graph evaluate every curve in parallel at the cost of one. Walrasian tâtonnement over encrypted demand and supply. How the engine finds a clearing price without decrypting a single order. ## Why this matters The matching engine's properties flow from the architecture. The order book stays encrypted because the engine doesn't need to decrypt it. The match is deterministic because the search is derived entirely from the encrypted inputs. The privacy is cryptographic rather than policy-based because there is no node, process, or human in the pipeline that holds the plaintext. Understanding the two parts — how the data is laid out and how the search runs on it — is sufficient to verify each of those claims independently. Both sections link to the interactive walkthrough at [dagon.so/how-it-works ↗](https://dagon.so/how-it-works) where each concept is illustrated step-by-step.