Skip to content

Privacy & Data

This page explains the practical data model users usually care about before they trust Fluent with real context.

The Short Version

Fluent stores the structured context your assistant needs in order to stop starting from scratch.

That can include:

  • profile and preference data
  • domain onboarding state
  • meal plans, recipes, and inventory
  • closet items and related media metadata
  • health preferences, goals, blocks, workout logs, and body metrics

Where Data Lives

Early Access

In early access, data lives in the managed Fluent environment.

Early access is currently free. It is the managed path, so the tradeoff is convenience and setup in exchange for using Fluent's managed storage and auth model.

You can start an export or deletion request in Fluent. During early access, the Fluent team completes and verifies export packaging, deletion, account closure, and migration requests.

Run Fluent Yourself

If you run Fluent yourself, data lives in your own environment.

By default that means ~/.fluent/, including:

  • local relational state
  • local artifact storage
  • local token state

If data location and operator control are your top priority, the open-source runtime is the clearest fit today.

What Fluent Stores

The exact data depends on which domains you use.

Shared Profile

  • display name
  • timezone
  • enabled and ready domains
  • onboarding state

Meals

  • preferences
  • meal plans
  • recipes
  • inventory
  • grocery plan state
  • meal feedback and related history

Style

  • style profile
  • closet items
  • item descriptors
  • provenance metadata
  • references to uploaded or attached media

Health

Health data supports limited early-access fitness and routine planning. It is not medical advice, diagnosis, treatment, emergency support, clinical nutrition, or a substitute for a professional.

  • training preferences
  • goals
  • training blocks
  • workout logs
  • body metrics
  • review context

Early Access Data Lifecycle

Retention

Early access data is retained for the life of the account, not just for a single chat or assistant session.

That means:

  • profile, domain, and workflow state remain available across sessions while the account is active
  • there is no user-facing retention timer or self-serve purge control in early access
  • if you want shorter retention windows or direct operator control, running Fluent yourself is the better fit today

Export

Early access exports use the same snapshot shape documented for the open-source runtime.

An export snapshot includes:

  • relational rows
  • artifact metadata

An export snapshot does not include:

  • artifact binary bytes

During early access, you can start an export request in Fluent. The Fluent team completes packaging, verification, and delivery rather than treating it as an instant self-serve download.

Start from meetfluent.app/account. Export requests and recent export status are available from the account path when your hosted account is active. If the account page is unavailable or your grant is stale, sign in again at meetfluent.app/sign-in or email [email protected] with the account email.

Deletion

During early access, you can start account deletion in Fluent. The Fluent team completes and verifies the deletion.

Practically, that means a deletion request should be treated as:

  • closing access for that account
  • removing managed relational state
  • removing managed artifact storage associated with that account

If you want a portable copy first, request an export before deletion.

Start deletion from meetfluent.app/account/delete. If you only submitted the waitlist form and never had an active Fluent account, use the deletion link in the confirmation email or contact support. Deletion normally requires operator verification during early access.

Account Lifecycle

The current account lifecycle is:

  1. You submit sign-up interest.
  2. Fluent grants access and provisions the account.
  3. You connect your assistant and use Fluent normally.
  4. If you need to leave early access, start the export or deletion request in Fluent, then the Fluent team completes and verifies it.

If an account is later re-enabled, expect to reconnect your assistant client and refresh any stale grants.

What About Photos And Files?

Fluent can store artifact metadata and related files depending on the workflow.

For example:

  • Style may use product images, fit photos, or other item evidence
  • early access stores artifact bytes in managed storage
  • self-hosted deployments store artifact bytes in your own storage backend

Across both paths:

  • snapshots move artifact metadata, but not the artifact bytes themselves
  • a JSON backup is not automatically a full binary-media export
  • moving artifact files between environments is still a separate workflow

Who Owns The Data When You Run Fluent Yourself?

Operationally, you do.

If you run Fluent yourself, you are responsible for:

  • where it runs
  • how it is backed up
  • who can reach it
  • reverse proxy and TLS setup if you expose it remotely

What Security Model Does The Open-Source Runtime Use?

The open-source runtime uses bearer-token auth on /mcp.

It does not try to be a full managed auth product. You should treat it as plain HTTP behind a trusted network or a properly configured reverse proxy.

See Configuration for deployment expectations.

What Do Backups Include?

Snapshot backups include:

  • relational rows
  • artifact metadata

Snapshot backups do not include:

  • artifact binary bytes

See Backup & Restore for the operator details.

How Migration Works

Relational data is designed to move through the shared snapshot format used by both paths.

Today:

  • early access to self-hosted is supported through snapshot export plus import
  • self-hosted backend moves are self-serve for relational rows and artifact metadata
  • self-hosted to early access migration should be coordinated with the Fluent team during early access

If the migration includes photos or files, artifact bytes need a separate transfer workflow in addition to the snapshot.

What Is Self-Serve Versus Operator-Assisted Today?

Self-Serve Today

  • running Fluent yourself
  • backing up and restoring self-hosted relational state
  • connecting after your early access account has already been provisioned
  • reconnecting your assistant client when a grant is stale
  • initiating an export request from the runtime
  • initiating account deletion from the runtime

Billing and account management are not handled inside ChatGPT. If paid managed access is introduced later, those paths stay on meetfluent.app through Stripe, and Fluent does not collect card details directly.

Operator-Assisted Today

  • granting early access
  • provisioning accounts
  • export packaging, verification, and delivery
  • deletion fulfillment, purge verification, and account closure
  • self-hosted to early access migration planning and import
  • artifact byte transfer between storage backends or hosting tracks

Should Privacy-Conscious Users Request Early Access Or Run Fluent Themselves?

If your main concern is maximum operator control, running Fluent yourself is the simplest answer today.

If your main concern is convenience and you are comfortable with managed storage and auth, request early access.

Fluent is in early access and open source