Platform Docs · v2.4

Build with encrypted compute.Documentation

Resources, guides, and API documentation for the LatticaAI platform

Encrypted query · live
Getting started

Two paths through the platform

Whether you’re shipping a model or querying one, the integration takes four steps. Choose your role.

Service Providers

Deploy
Account

Create your account

Sign up at the Lattica console and set up your organization.

Deploy

Deploy your workload

Upload your workload and the platform compiles it for encrypted execution.

Configure

Buy credits

Purchase credits to make your workload available for queries.

Monitor

Monitor usage

Track the number of queries made against your workloads.

End Users

Query
Install

Install the client

Download the Lattica's Query Client.

Authorize

Get an access token

Request a token from the workload provider you want to query.

Encrypt

Encrypt your input

Your data is encrypted on your device before it's sent. It stays private throughout.

Decrypt

Get your result

Receive the encrypted output and decrypt it locally with your client.

Key concepts

The vocabulary of encrypted compute

Four ideas you'll see across every guide. Internalize these and the rest of the docs will read fast.

FHE

Fully Homomorphic Encryption

Compute on ciphertext. Models perform inference directly on encrypted inputs and produce encrypted outputs - no decryption step on the server side, ever.

HEAL

Homomorphic Encryption Abstraction Layer

HEAL bridges the gap between FHE software and hardware, enabling integration of acceleration hardware with Lattica’s FHE-as-a-Service platform.

SDK

Client SDKs

SDKs for every role: query users encrypt and submit queries, workload providers manage deployments, and developers build homomorphic pipelines directly.

EEL

Encrypted Execution Layer

The runtime where encrypted queries are processed. Sees only ciphertext.