An AI Workspace Instead Of Scattered Chats
An AI workspace is a single place where your conversations, model choice, uploaded files and reference documents live together. ChatOTL persists all four per account, and exposes them to agent clients over an OAuth-protected MCP server.
A single chat window is a scratchpad. A workspace is what you get when the thread history, the files you attached, the reference documents you curated and the prompts you reuse are all still there tomorrow.
What the workspace holds
- Conversations with full history, keyword search and branching.
- A per-message model choice across several providers.
- Uploaded files, stored privately and read through short-lived signed URLs.
- Knowledge bases: your own documents, chunked and embedded so answers can draw on them.
- Reusable system prompts you can apply when starting a new thread.
Agent access over MCP
The workspace is also reachable from agent clients. An OAuth-protected MCP server lets Claude, Cursor or ChatGPT list, read, search, rename and create your conversations with your permission.
Where the boundaries are
Everything in the workspace is scoped to your account. Nothing is shared with other accounts, and there is no cross-account team view.
Limitations
- Single-account workspaces: no shared team spaces or per-seat collaboration yet.
- No live web browsing inside the workspace.
- Knowledge-base answers depend on the documents you supply and still need review.
Frequently asked questions
What makes this a workspace rather than a chat app?
Conversations, files, knowledge bases and reusable prompts persist and are searchable, and the same data is reachable over MCP.
Can teammates share a workspace?
Not currently. Each workspace belongs to one account.
Does the workspace support MCP clients?
Yes, through an OAuth-protected MCP server documented in the MCP setup guide.
Set up your workspace
Create an account, pick a model, attach your first document, and connect an MCP client when you're ready.
Create an account