boxes.dev vs Replicas
Both give agents isolated cloud machines and use your Claude and ChatGPT subscriptions. Replicas builds workspaces from hook scripts for Slack triage and event automations, and boxes.dev runs each task on a full, isolated copy of your development environment.
The short answer
Replicas runs Claude Code, Codex, and six other agents in cloud VMs built by hook scripts, and starts work from Slack, Linear, GitHub, automations, and an API. boxes.dev runs Claude Code and Codex on full, isolated copies of a cloud machine where your app already runs. Choose Replicas for more agents, Slack channel triage, or automations on GitHub and Sentry events. Choose boxes.dev when you want the agent to test its work in your full running app, like an engineer would, your environment is easier to copy than to script, or you want machines that stay until you delete them.
Choose boxes.dev if
- You want the agent to test its changes in your running app, with its real databases, services, and data, the way an engineer would.
- Your environment is easier to copy from a machine where your app already runs than to script with hooks.
- You want Claude Code and Codex in both a graphical view and their own terminal UI.
- You want each task's machine to stay until you delete it, with dev servers and databases resuming when it wakes.
- You need one stable public URL for OAuth callbacks or webhooks that you can point at any devbox.
Choose Replicas if
- You want agents beyond Claude Code and Codex, such as Cursor, OpenCode, or Pi.
- You want an agent that watches Slack channels and decides which messages need work.
- You want automations started by GitHub events or Sentry alerts.
- You want agents to build and test iOS and Android apps on cloud simulators and emulators.
- You want unlimited workspace time for the work each person starts, on a Team seat.
At a glance
As of September 2026.
| boxes.dev | Replicas | |
|---|---|---|
| Agents | Claude Code and Codex, each with your own plan or an API key, in a graphical view or the agent's own terminal UI | Claude Code, Codex, Cursor, OpenCode, Pi, and three other agents, each with your own subscription, an API key, or a cloud model account, in Replicas's chat view |
| Where cloud work runs | A devbox for each task: a full, isolated copy of your Template box that can hold several threads | A VM for each task, called a workspace, that can hold several chats |
| Cloud environment | A snapshot of your Template box, a cloud machine that a setup agent prepares with your app's tools, databases, services, and data and a maintenance script keeps current; a new devbox can be prepared while you type, and a startup script runs on it | A snapshot that warm hooks build, kept as a pool of ready workspaces; start hooks run at boot |
| When a cloud machine goes idle | It sleeps after about five minutes without use, keeping files and memory, so dev servers, databases, and terminals resume when it wakes | It sleeps after an hour without activity, keeping its desktop and processes |
| After the work is done | A devbox stays until you delete it | A workspace is archived after 7 days asleep and deleted after 30 more days without activity |
| Working in the cloud machine | Terminal tabs, a file editor, Git history and diff review with line comments, a browser you share with the agent, previews of images, PDFs, and Markdown, a process monitor, and SSH | Terminals, a file editor, changed files with line comments, a streamed desktop with Chrome, Canvas for files the agent makes, and SSH |
| Reaching the running app | A browser on the devbox that you and the agent share, ports forwarded to the same port on your laptop, stable .localhost addresses, public links, and Project URLs | Chrome on the workspace's streamed desktop, tunnels to the same port on your computer, and public preview links, optionally behind a Replicas login |
| Apps | macOS desktop app, iOS and Android apps with a terminal, the dvb CLI, SSH, and VS Code or JetBrains through dvb ide | Web, a desktop app, an iPhone app, the replicas CLI, SSH, and VS Code through replicas code |
| Starting work | The apps, dvb claude or dvb codex from your terminal, @boxes.dev in Slack, assigned Linear issues, and automations on a schedule or from an API call | The apps, replicas create from your terminal, @Replicas in Slack and watched channels, assigned Linear issues, GitHub mentions, automations on a schedule, from an API call or webhook, or from GitHub, Sentry, or Slack events, and MCP |
| Pull requests | Pull requests can fix failed checks and bot review feedback on their own, and merge from Git/Review | Failed CI checks and review-bot comments go to a separate chat that pushes fixes, and eligible pull requests merge from the workspace |
| Shared setup | Environment files with project, team, and personal values; startup and maintenance scripts; skills, MCP servers, and agent plugins on every devbox; AWS, Tailscale, and a static egress IP; Team Templates | Global, team, and personal environments with variables, files, skills, MCP servers, and plugins; Doppler or Infisical secrets; a static egress IP and short-lived identity tokens on Team |
| Working with your team | Read-only thread links, team-wide thread reading, and project browsing for teammates; @boxes.dev in Slack, where teammates you allow can start work and send follow-ups that run with your access; agents that read earlier threads; and commits linked to threads | Workspaces that every member can open, using the creator's or the organization's agent credentials; @Replicas follow-ups in the same Slack thread; contributor lists on shared workspaces; and long-term memory that carries lessons into later work |
| Pricing | Free trial with 10 box-hours; Starter $19 and Pro $99 per user each month, plus your Claude or ChatGPT plan or an API key | 14-day free trial; Developer $50 and Team $200 per Full seat each month, or Flex seats by the minute; API and automations billed by the minute |
How environments are made
Replicas builds each workspace from environment settings and hook scripts, and boxes.dev copies a machine where your app already runs.
A Replicas organization has a Global environment plus team and personal
environments. Warm hooks, set in the app or in a replicas.yaml file in the
repository, install dependencies and run builds ahead of time. Replicas
snapshots the result and keeps a pool of ready workspaces, five by default for
a team environment. Start hooks run at boot for work such as starting dev
servers, applying migrations, or pulling current data. Replicas bills workspace
time by the minute while a workspace is awake, and a Full seat covers the work
that person starts, up to 150 hours a month on Developer.
boxes.dev first sets up your full development environment on a cloud machine called the Template box, with the tools, databases, services, and data your app needs, so an agent can run the app and test its changes end to end. A setup agent can port the environment from your laptop, including the files, local data, agent settings, and skills you choose to upload, or build it from a GitHub repository. boxes.dev then saves the Template box as a snapshot, and every task runs on its own devbox, a full, isolated copy of that machine. A maintenance script can keep the Template box current as your code changes, running on a schedule or when files such as a lockfile change, and each successful run saves a new snapshot. Team Templates let an admin prepare one starting environment for everyone on the team.
You'd never let an engineer ship code they hadn't run, and a full environment holds the agent to the same standard. On a boxes.dev devbox, the agent runs your whole app, with its databases, services, and data, and tests its changes in the browser as it builds. It reproduces bugs in the real app to find the root cause instead of guessing from the code, so it writes better code and makes fewer mistakes. It can also take screenshots and record video for you to review, so you can hand it most repetitive manual testing.
Reaching the running app
In both products, you and the agent can use the app running in the cloud, and boxes.dev also gives your project one address that you can move between devboxes.
Every Replicas workspace has a streamed desktop with Chrome that you can watch or take over. You or the agent can register a workspace port as a public preview link, optionally behind a Replicas login, and detached services keep dev servers running across turns and sleep. The CLI and the desktop app can tunnel a workspace port to the same port on your computer.
Each boxes.dev devbox runs a browser that you and the agent share, so you can
watch the agent test the app, click through it yourself, or comment on page
elements to show the agent what to change. Ports forwards the app to the same
localhost port on your laptop, a public link shares it with anyone, and a
Project URL gives your project one stable public address that you can point at
any devbox, for OAuth callbacks and webhooks.
Working with your team
Both take requests in Slack threads, and Replicas can also watch whole channels. Every Replicas member can open any workspace, and on boxes.dev, teammates read your threads, while only teammates you allow can send your agent work from Slack.
Every member of a Replicas organization can open any workspace, and shared
workspaces list their contributors. A workspace uses its creator's
coding-agent credentials, or the organization's when the creator has none, and
Replicas's docs describe members as trusted users who can retrieve the
organization's shared credentials through the agent or terminal. Pull requests
come from the Replicas bot unless you turn on attribution, which uses the
requester's own code-host login. Mentioning @Replicas in Slack starts a
workspace, and later mentions in that thread go to the same workspace.
Replicas's Mothership agent can also watch the channels you choose and decide
which messages need a reply or a workspace. Long-term memory carries team
decisions and lessons from past work into later workspaces.
Each boxes.dev devbox belongs to one person and works with their access, such as their GitHub login and cloud credentials. Anyone who can prompt its agent could act with that access, so in the app, teammates get read-only access to your work. You can share a thread as a link, and a team admin can let teammates read every thread that isn't marked private and browse each other's projects in a Team workbench. Teammates can't send prompts, open your terminal, or control your devbox.
In Slack, mentioning @boxes.dev in a thread starts Codex or Claude Code on a
new devbox, and later mentions in that thread send follow-ups to the same
conversation. By default, only you can start work in your project from Slack
or follow up on it. If you allow teammates, they can also start work there
with @boxes.dev @YourName and send follow-ups, and their requests run with
your access and usage. boxes.dev sends you a DM with links to each task a
teammate starts. Agents can read earlier threads, and commits can link to the
thread that produced them.
What comes with every boxes.dev devbox
The boxes.dev desktop app opens the same set of tools for whichever devbox you select:
- Terminal, with tabs that keep running after you close the app.
- Files, a file browser and editor with search across the workspace.
- Git/Review, for branch history and diffs, where comments on changed lines go to the agent together as one message.
- Browser, the browser on the devbox that you and the agent share.
- Ports, for forwarding to your laptop, stable
.localhostaddresses, and public links. - Activity, which shows the processes using the machine's CPU and memory.
Agents can start other agents on new or existing devboxes, message each other,
and leave notes in a shared /project folder, and you can branch a
conversation onto a new devbox with the current files.
Where Replicas shines
- Eight agents. Claude Code, Codex, Cursor, OpenCode, and others, with chats that can move to another agent and keep their context.
- Slack channel triage. Mothership watches the channels you choose, decides which messages need a reply or a workspace, and follows instructions you set for each channel.
- Automations on outside events. GitHub events, Sentry alerts, and Slack messages can start fresh workspaces, and a run can report its verdict as a GitHub check.
- Mobile app testing. Agents can build and test iOS and Android apps on cloud simulators and emulators from a workspace.
Pricing
As of September 2026.
| boxes.dev | Replicas | |
|---|---|---|
| Price | Trial free; Starter $19 and Pro $99 per user each month | 14-day free trial; Developer $50 and Team $200 per Full seat each month; Flex seats pay by the minute |
| Model usage | Your Claude or ChatGPT plan, or an API key | Your subscriptions, API keys, or cloud model accounts |
| Cloud time included | 10 box-hours on the Trial, 40 per user on Starter, 250 on Pro | 150 hours per Developer Full seat and no cap on Team, for work people start; none for API and automations |
| Cloud machines awake at once | 2 on the Trial, 4 per user on Starter, 10 on Pro | 50 manual workspaces on the trial; limits for paid plans aren't published |
| Extra cloud time | $0.75 per box-hour on Starter, $0.60 on Pro | Flex seats at $0.48 an hour on Developer and $0.96 on Team; API and automations at $0.48 an hour for 2 CPUs and 8 GB, or $0.96 for 4 CPUs and 16 GB |
A box-hour is one default devbox awake for an hour, and sleeping devboxes don't use any. On Starter, 40 box-hours works out to about two hours of devbox time each working day, and Pro's 250 to about twelve, spread across as many as ten devboxes at once. Your Claude or ChatGPT plan covers the agents' usage. See Plans, seats, and box-hours.
A Developer Full seat's 150 hours cover the workspaces that person starts, whatever their size, and two workspaces running for an hour use two hours. At 150 hours, that seat's workspaces pause until the next billing period. Work that the API or automations start is billed to the organization by the minute. See Replicas's pricing and billing docs.
Using both
Replicas and boxes.dev can work on the same repositories and pull requests. You can keep Replicas for Slack channel triage and event-driven automations, and use boxes.dev when a task needs Claude Code or Codex on a copy of your full running environment.
Frequently asked questions
Can I use my Claude and ChatGPT subscriptions with both?
Yes. Replicas connects a Claude subscription with replicas claude-auth and
supports ChatGPT sign-in for Codex. boxes.dev runs Claude Code with your
Claude plan and Codex with your ChatGPT plan, or with an API key.
What's the difference between a Replicas workspace and a boxes.dev devbox?
Both are isolated cloud machines that sleep when idle and keep their running processes. A Replicas workspace starts from an environment's warm snapshot and is archived, then deleted, after long inactivity. A boxes.dev devbox starts as a copy of your Template box and stays until you delete it.
Can I open a devbox in my own editor?
Yes. dvb ide opens a devbox in VS Code or a JetBrains IDE over SSH. See
Install and use the CLI. Replicas can open a workspace in VS
Code through replicas code.
Sources
- Replicas: overview, workspaces, environments, repository config, Canvas, chat forking, long-term memory, workspace identity, static egress IP, credentials, Claude Code, Codex, CLI, desktop app, iOS app, Slack, Linear, GitHub, automations, MCP, mobile testing, team, billing, changelog, and pricing.
- boxes.dev: How boxes.dev works, Update your Template box and snapshots, Startup, teardown, and maintenance, Sleep, wake, and recover a devbox, Review a page with your agent, Preview your app, Project URLs, Browse, edit, and transfer files, Create and publish a Team Template, Connect Codex and Claude, Use the agent's terminal view, Environment files and secrets, Add MCP servers, Install agent plugins, Organize and share threads, Team access and roles, Branch a conversation, Shared project storage, Commit, push, and open a pull request, Install and use the CLI, Work from Slack, Work from Linear, Create an automation, Use boxes.dev on your phone, and Plans, seats, and box-hours.
Product names are trademarks of their respective owners.
