Plan here, build there
Think in Claude, implement in Cursor, draft in ChatGPT — and carry the brief between them instead of rebuilding it.
Save a block of context in one assistant and get a short handle back. Load it by that handle in another — no copy-paste, no re-explaining. Your context, carried across tools.
You plan in one assistant, build in another, and draft in a third. Each one holds real working state — the brief, the constraints, the decisions so far. None of them will hand it to the next.
So the state of your work gets rebuilt from memory, every time you switch tools. That tax is small enough to pay once and large enough to feel every day.
Retype the background in every new tool, losing fidelity each time.
Haul transcripts between windows. Clumsy on a desktop, hopeless on a phone.
Each assistant remembers — for itself. It can't brief a competitor's tool.
Connect any MCP-capable AI to Segue over OAuth — then let it carry context for you.
Handles are three pronounceable syllables. Say one aloud, type it on a phone, write it on a sticky note.
One AI calls save_context with your working brief, notes, or plan — and gets back a short handle like brelavo.
Hand that handle to another assistant — anything connected to the same Segue account.
The next AI calls load_context with the handle and picks up exactly where you left off.
Think in Claude, implement in Cursor, draft in ChatGPT — and carry the brief between them instead of rebuilding it.
Hitting a context limit or closing for the night? Save a curated brief instead of dragging a sprawling transcript forward.
Keep project background, house style, and preferences saved — and bootstrap any newly connected assistant on demand.
Segue is a small, neutral relay between your AIs — built on commitments, not features.
Nothing syncs behind your back. An AI saves, you pass the handle, another AI loads. Every transfer is one you chose.
The unit of exchange is a word you can say aloud or type from memory — even when the context behind it runs to 100,000 characters.
Segue belongs to no assistant. Any MCP-compliant client connects the same way — today's and whatever you switch to next.
A handle resolves only inside your account. Leaked or guessed, it's meaningless to anyone else — by construction.
Every context is readable text you can browse, edit, search, and download as a .md file. No lock-in format.
Each AI gets scoped permissions and short-lived tokens, listed in Settings with one-click revoke. Stored credentials are hashed.
Your data is never hostage. Loading, editing, exporting, and deleting stay free on every plan, in every state — lapsed ones included. Paying buys headroom, never access to your own words.
Three pronounceable syllables — like brelavo — minted when a context is saved. It's the context's address: case-insensitive, unique, and resolvable only inside your account.
Any client that speaks MCP with streamable HTTP and OAuth — Claude (Desktop, Code), ChatGPT, Cursor, Windsurf, Codex, and more as the ecosystem grows. Setup is pasting one prompt: the AI configures itself and walks you through sign-in.
The Model Context Protocol — an open standard, stewarded by the Linux Foundation, that lets AI assistants connect to outside tools. Segue is one server that all your assistants share.
No. Handles only resolve for clients signed in to your account. Outside it, a handle is just a made-up word.
Never. Nothing is saved unless an AI you connected calls save_context in front of you, and nothing loads unless you pass the handle. No ambient memory, no surprises.
Your account drops to the free tier. Everything already saved stays loadable, editable, exportable, and deletable — only new saves pause while you're over the free cap.
Create a free account, connect your AI over MCP, and start handing off context.
Start for free