Overview

We've built Prefix with great care around what data gets sent to AI models, how secrets are handled, and where actions are executed from.

Currently, with the sole exception of the AI model, Prefix can be entirely self-hosted. Provided with a private cloud AI deployment, we are able to meet most standards for internal tooling at larger organizations.

Data Handling

Each action on Prefix allows you to configure exactly what metadata from the action gets sent to the AI for automation. This allows you to import actions and build automations using test data, but run them using production data, in a way where the data you care about never even leaves your servers.

We use OpenAI's chat completion API endpoint for much of our AI work currently. OpenAI has evolving data usage policies, but provides clear guarantees data sent through this endpoint is never used for training purposes in ways which could create exposure. Read More (opens new window)

Internal Resources

Self-hosting also means Prefix can access and help you automate publicly inaccessible internal resources, with very little overhead.

Our terraform script can setup a private deployment of Prefix on AWS from scratch in less than 5 minutes. We keep starter deployments extremely simple, requiring only require a single t4g.medium EC2 instance.