Roy PadinaMac apps
AgentsSessionHopper

AgentsSessionHopper

BETA: not fully tested; real account hand-offs are new.

Profiles for Claude Code. One conversation, any of your accounts.

Keep your own Claude accounts side by side, one Claude Code profile each. When one runs low, continue the same conversation on another with one click. Nothing lost, nothing re-explained, and nothing moves until you say so.

brew install --cask roypadina/tap/agents-session-hopper

Version 0.1.6, released 24 October 2026. macOS 14 or later, Apple Silicon and Intel. Free and MIT licensed.

GitHubWiki
The AgentsSessionHopper menu-bar window: three profiles with 5-hour and weekly usage, a prompt to move a conversation from work to personal at turn end, live sessions, and recent hand-offs.

BETA: not fully tested; real account hand-offs are new

The engine and the CLI are covered by unit and integration tests against a fake claude, and the app has been used day to day on one Mac. Hand-offs between real accounts are the new part. Expect rough edges, read the "what it never does" list below before you rely on it, and report what breaks.

Why AgentsSessionHopper

  • One conversation, any of your accounts

    When the account you are working on runs low and another has room, you are asked whether to continue there. Confirm, and the session moves when the current turn ends, resuming the same conversation under the next profile. Nothing re-explained, nothing lost.

  • You stay in control

    The suggestion comes at 85 percent, adjustable. Moving without asking is off by default and explained in Settings. Past 97 percent with nothing confirmed, the session is parked at its next turn end instead of running into paid extra usage, and tells you how to resume it. Every suggestion, move, park and limit stop is in a 90-day log.

  • Menu bar, terminal, or both

    The menu-bar window shows each profile's 5-hour and weekly usage, live sessions and recent hand-offs. The sessionhopper CLI does all of it from the terminal, and a shim lets cmux or an IDE start every claude supervised.

How it works

Profiles, usage, a move.

A profile reaches 85 percenta notification and the menu bar ask; nothing moves until you confirm
You confirmthe turn and any background jobs finish, then claude --resume restarts under the next profile
Claude Code stops on a limityou are asked whether to continue elsewhere; the work picks up where it stopped

A profile is a Claude Code configuration folder. Existing ~/.claude and ~/.claude-* folders are picked up; new ones are created with your setup and history shared, then logged in through Claude Code itself. sessionhopper starts the unmodified claude with that folder and per-session settings that add a status line and a few hooks; no settings file is written. Moving needs every profile to share one projects/ folder, where Claude Code keeps conversations; profiles the app adds do, and sessionhopper doctor checks.

Sibling app: Agents Monitor watches usage limits for Claude Code and Codex accounts from the menu bar, polling the providers' own usage endpoints. AgentsSessionHopper reads nothing from the network and only moves conversations. They are independent and work side by side.

The window and the settings

Settings, Hand-off: suggest a move at 85 percent, move without asking switched off, park before paid usage at 97 percent, hold every session.
Settings, Hand-off: the suggestion threshold, move without asking (off), park before paid usage, hold.
Settings, Profiles: the fallback order of three profiles with their configuration folders, Add Profile and Check logins buttons.
Settings, Profiles: the fallback order; new sessions start on the first profile with room.
Settings, Log: a table of hand-off events with time, event, from and to profile, reason and folder.
Settings, Log: every suggestion, move, park and limit stop, with the reason and the folder.
Settings, Command line: where the sessionhopper CLI lives, the alias to supervise every claude, and the shim link for cmux and IDEs.
Settings, Command line: the CLI, the alias to supervise every claude, the shim for cmux and IDEs.
The first-run guide: the beta warning, what the app never does, the three profiles found, and the not-affiliated-with-Anthropic note.
The first-run guide: the beta warning, what the app never does, and your profiles.

Requirements

  • macOS 14 or laterSonoma onwards.
  • Apple Silicon and IntelOne universal download; the CLI is bundled and linked by Homebrew.
  • Claude CodeInstalled, with a Pro or Max plan per account: usage numbers come only for those. Team and Enterprise seats can be profiles, but report no usage windows.
  • No network, no analyticsState lives in Application Support (profile list, last-known usage, the switch log). Open source under MIT.

First launch

AgentsSessionHopper is not notarized by Apple, which needs a paid developer account (it is ad-hoc signed). So macOS blocks the first launch with a "could not verify" message. Open System Settings > Privacy & Security, scroll down and click Open Anyway. Or clear the quarantine flag once:

xattr -dr com.apple.quarantine "/Applications/AgentsSessionHopper.app"

After that it opens like any other app. If you would rather not trust a download, build it from source.

Clearing quarantine also unblocks the bundled sessionhopper command.

What it never does

This tool sits next to your Claude Code accounts, so the boundaries matter more than the features.

  • It never touches your login

    It never reads, copies, stores or sends your Claude login, tokens or Keychain items. Login always happens in Claude Code's own claude auth login flow, one configuration folder per account, the way Claude Code documents running several accounts.

  • It makes no network requests

    None: no usage endpoint, no telemetry, no update check. Usage comes only from the documented status-line input Claude Code gives its own status line, and limit stops from the documented StopFailure hook.

  • It does not change or get around any limit

    Each account keeps its own limits. You decide when a conversation continues elsewhere, and nothing moves until you confirm, unless you switch on "Move without asking" yourself. Claude Code's traffic is never changed, proxied or inspected; it runs the official, unmodified claude.

Your part: use it only with accounts that are yours, never shared with anyone. Each account is subject to Anthropic's Consumer Terms, Usage Policy and Claude Code usage rules, which Anthropic may change at any time.

Not affiliated. AgentsSessionHopper is an independent open-source tool. It is not made by, affiliated with, or endorsed by Anthropic. Claude and Claude Code are trademarks of Anthropic, PBC.

Questions

Does it read my usage from Anthropic's servers?

No. It makes no network requests. It only uses the numbers Claude Code gives its own status line, so a profile with no running session shows "unknown".

Does it move sessions on its own?

Only if you turn on Move without asking, which is off by default. Otherwise every move is your click or command. Park before paid usage only ever ends a session; it never moves one.

Can I use it with someone else's account?

No. Use it only with accounts that are yours, never shared. Each account stays subject to Anthropic's terms.

Does it work without the app?

Yes. The CLI does everything; the app adds the menu bar, notifications and settings.

Codex?

Not yet. Claude Code only.

Keep the next one coming

If AgentsSessionHopper has saved you from re-explaining a task on another account, you can support its development. Optional, always appreciated.

I'm a software engineer from Israel who builds small, focused Mac tools to fix the little annoyances in my own day — then shares them free and open source. If this app saves you time, a coffee on Ko-fi keeps the next one coming. ☕

Roy Padina

Buy me a coffee on Ko-fi

Not today? A star on GitHub or a bug report helps too.

What a coffee pays for

  • The Apple Developer Program, so the apps can be notarized and the "Open Anyway" step goes away.
  • Evenings on bug reports, new macOS versions and Intel builds.
  • The next small tool.

Source, docs, bugs