Stopover Cloud
Private alpha

Switch agents. Pick up where you left off.

Carry your goal, decisions and next steps between Claude Code, Codex, Cursor and any MCP-compatible agent. Say “save a handoff” in one, then “resume” in another—even on a different machine.

Free during the alpha. Sign in to join the waitlist.

Claude Code · ~/api

Save a handoff so I can continue this later.

stopover_save · “Fix refresh-token rotation”

Saved. 2 of 3 steps done, tests next.
Resume with Resume Handoff 3f2a9c1e

Usage limit reached · resets in 3 h

Codex · ~/api

Resume Handoff 3f2a9c1e

stopover_resume · “Fix refresh-token rotation”

Picking up “Fix refresh-token rotation”. Rotation and reuse detection are done; you chose 30-day expiry for mobile. Writing the expiry tests now.

Works with

How it works

  1. 01

    Connect each agent once

    Add one URL to each agent. It opens your browser to sign in with GitHub. No token to copy.

    claude mcp add --transport http stopover-cloud \
      https://stopover.cloud/mcp
  2. 02

    Save when you switch

    Ask in plain words. The agent writes the summary: objective, decisions and why, progress, dead ends, next steps.

    > Save a handoff so I can
      continue this later.
  3. 03

    Resume anywhere

    Paste the prompt into any connected agent, on any machine. It claims the task so nobody else picks it up.

    > Resume Handoff 3f2a9c1e

Your dashboard

Every handoff and memory, in one place.

See what’s ready to resume, what an agent is working on and what’s done. Open a handoff to read its objective, decisions and next steps, or browse what your agents remember about you.

Why Stopover Cloud

Built for work that outlives one session.

One inbox for every agent

Handoffs aren’t tied to a vendor. Save in Claude Code, resume in Codex, check in Cursor. Your agents use the same tools and prompts everywhere: save, resume, inbox.

No double work

Resuming claims the task. A second agent sees it’s taken, and an expired claim can be picked up safely.

Done means done

An agent can only mark a task succeeded with evidence for every acceptance criterion.

A memory that follows you

“Update my Stopover memory” in one agent, “import my Stopover memory” in another. Preferences, project context and corrections travel with you, with revisions so agents never overwrite each other, and forgetting reaches every agent.

Only you can see it

Your handoffs and memories belong to your account alone. Revoke any agent’s access at any time.

Safe to retry

Dropped connection mid-save? Agents retry with the same operation ID, so nothing is saved twice.

Open core

Built on Stopover, the open-source inbox.

Stopover Cloud is the hosted edition of Stopover (Apache-2.0). Run Stopover yourself in one Docker container, or let Stopover Cloud handle hosting, backups, sign-in and access from anywhere.

Questions

What exactly is a handoff?

A summary your agent writes when you ask it to: the objective, background, requirements, decisions and their reasons, progress, failed attempts, open questions, references and next steps, plus acceptance criteria. You don’t fill in fields or write JSON.

Does Stopover Cloud copy my code or files?

No. It carries context, not the workspace. The agent that resumes needs its own access to the repository and tools, and checks your current instructions before it continues.

Which agents work?

Any client that supports MCP over Streamable HTTP: Claude Code, Codex, Cursor, Hermes Agent, Oh My Pi and others. Agents that support MCP sign-in connect with just the URL; for the rest, create an access token.

What does it cost?

Nothing during the alpha. Pricing will be announced before any paid plan starts, and there will be a free tier.

Who can see my handoffs?

Only you. Each workspace is isolated, and access tokens can be revoked at any time. Anything an agent reads can reach that agent’s model provider, so keep secrets out of handoffs.

Stop re-explaining your work to every agent.