Fluent Case Study

Why I Built Fluent

AI assistants are often weakest at the ordinary details that make advice useful. Fluent gives those details a portable, source-aware place to live.

The Product Thesis

Fluent is a personal-context and action layer beneath the assistant a person already uses.

It is designed around several questions:

• Is this fact confirmed, inferred, or stale?
• Where did it come from?
• Should this AI app be allowed to use it?
• Can the assistant act on it?
• Did the change actually save?
• Can the user export or delete it?
• Can the same product work as both a managed service and open-source software?

Current Product

• Meals turns household context into an assistant-drafted plan, an approved save, a living grocery list, and useful feedback after cooking or shopping.
• Style connects closet items, photos, purchase questions, and approved closet updates.
• Budgets adds simple grocery and clothing amounts to Meals and Style decisions.

Architecture

Fluent uses one MCP interface across ChatGPT, Claude, Codex, OpenClaw, and other compatible AI apps.

The main design choices are:

• typed shared and domain context
• a record of where important details came from and how certain they are
• tenant-aware hosted storage
• approval before specific changes are saved
• verification that saved changes worked
• optional visual views that show the same saved information
• a shared contract across managed and open-source runtimes

Fluent always works in text. Compatible apps can also show a visual Grocery List, Budget Setup, or Style Closet Manager.

Availability

Managed Fluent is currently free.

The open-source version is available for anyone who wants to inspect or run Fluent themselves.

What This Project Demonstrates

Fluent is an exercise in product architecture as much as implementation: deciding what an AI assistant should know, what it may change, what it must prove, and how those rules remain consistent across AI apps and setup options.