Multi-Model

Your AI context belongs to you — not to any provider. Switch models at any time without losing a thing.

Freedom From Vendor Lock-In

Every other AI tool traps your work inside its own walls. Months of conversation with ChatGPT. History, artifacts, context, decisions. All of it locked to one provider. If a better model comes along — or if you just want to try something different — you start from scratch.

LIT breaks this. Your history, your artifacts, your project context all live in LIT — stored locally, on your infrastructure. The AI provider is just a backend. Switch from ChatGPT to Claude. Try Gemini for a week. Come back. Nothing is lost. The channel is still there, the messages are still there, the work is still there.

This is a different relationship with AI. You're not a tenant in someone else's system — you're the owner of your own context.

Switch Anytime, Pick Up Immediately

Switching models is as simple as changing the model selector. The next message picks up exactly where you left off — with full access to the channel history, uploaded files, and all prior work.

The new model can read everything the old model did. It can reference artifacts, continue partially completed work, and build on prior decisions. You don't re-brief it. You don't re-upload files. You don't summarize the last three months of work. You just keep going.

Route by Strength

Different models are better at different things. Use Claude for reasoning-heavy work and code. Use a fast, cheap local model for high-volume mechanical tasks. Use Gemini when you need multimodal capability. Use whichever frontier model is currently best for your domain.

Because context lives in LIT rather than the model, you can route each task to the right tool without paying a context tax every time you switch.

Multiple Models in a Channel

LIT is designed for AI to be a genuine participant in your work — which means more than one AI can be in the room at once. Agents can be assigned roles and specialties: a marketing-focused agent, an engineering agent, a research agent, each configured with a different model, system prompt, and tool set.

By default, one agent owns a channel and responds to messages. Others can be brought into the conversation with an @mention — @claude for a deep reasoning take, @gemini for a multimodal perspective. Each responds in context, with full access to the channel history.

The result is less like querying a tool and more like having a team.