Memory now travels with you between models
Tell TunnelGPT something once and every model knows it — your stack, your tone, your projects. You stay in control of every line.
Every AI chat starts with the same five minutes: re-explaining who you are, what you’re working on and how you like your answers. Multiply that by every model you use and it adds up fast.
Starting today, beta members get Memory — one shared memory that every model in TunnelGPT can read.
What Memory remembers
- You: your role, your expertise, the languages you write in.
- Preferences: “keep it concise”, “use British spelling”, “TypeScript, not JavaScript”.
- Projects: names, deadlines and context you’ve shared across chats.
You’re in control
Memory is a plain list you can read and edit at any time. Delete a line and it’s gone everywhere. Turn Memory off for a single chat with one click, or start a temporary chat that leaves no trace at all.
Models only receive the memories relevant to the current conversation, never the whole list — and memories are never used to train any model.
Why it matters more with many models
With a single model, memory is a convenience. With forty, it’s what makes switching feel effortless. Jump from a reasoning model to a fast one mid-thread and it already knows your project, your constraints and your style.
Every model in one place.
Reserve your workspace in early access.