**Built my own model-agnostic AI workstation because I was tired of platform lock-in — free, BYOAK, open source**

Tired of rebuilding context every time I switched models. Tired of my personas living inside OpenAI's walled garden. Built something to fix it.

**Architect's Domain**, a workstation UI that sits on top of any provider.

Core features:

- **Workspace system**, persistent environments with pinned context, imported files, notes. Think Claude Projects but provider-agnostic

- **Manual memory curation**, fragments surface during chat, you approve or reject what gets remembered. No silent auto-memory

- **Character/persona system via file injection**, load .txt files as system context. Works with character cards, lorebooks, personality files, anything

- **Provider switching**, OpenRouter, Venice.ai, DeepSeek. Swap models without losing your setup

- **BYOAK**, your keys, your data, runs fully static

No React, no framework bloat. Vanilla JS + CSS + HTML. Deployable anywhere.

I use it daily for prompt engineering and RP character testing across different frontier models. The workspace + memory combo is what makes it actually useful vs just another chat wrapper.

Open source: https://github.com/HactoriXD/architects-domainv1

Feedback welcome! especially from people who've tried similar setups.