Keep your workflow
Same CLI, same session. cerver swaps the harness underneath — you don't relearn anything.
01 — The trap
You picked your agent six months ago. A better one shipped in month two.
Switching costs feel high, so you don't — and you quietly pay for it on every task. The fix isn't loyalty to one model; it's a workflow that checks, every time, and stays on whatever's winning right now.
02 — Compare in one command
Run the same task on two agents. Keep whichever won.
One session, one prompt, a real diff — not vibes. The comparison runs itself; you just stay on the winner.
$ cerver run --compare claude codex "add idempotency to the webhook"
claude / sonnet passes · retries on 5xx
codex / gpt-5 passes · cleaner backoff ← better
→ you're now on codex for this task.03 — Switch without losing the thread
Swap the model or the machine mid-run. The transcript stays bound.
Keep the context
Transcript, tools, and identity travel with the session across every swap.
Run it anywhere
Your machine, Vercel, or E2B — move compute mid-task without dropping a beat.
04 — Install
One line. Free at hobby scale.
$
curl -fsSL https://cerver.ai/install.sh | bash
Bring your own compute. cerver charges nothing — your bill is whatever your provider charges, usually $0 at hobby scale.