If your team uses AI coding agents, you'll know the pull between running them locally, where you have control, and running them in the cloud, where it's convenient. On 23 September 2026 Anthropic moved Claude Code cloud sessions out of research preview and into general availability for Pro, Max, Team and Enterprise users with premium seats. It changes where the coding agent actually runs, and that has knock-on effects for collaboration, security and day-to-day development.
Until now Claude Code mostly lived in a developer's terminal, tied to their machine. Cloud sessions run in isolated sandboxes on Anthropic's infrastructure and can be reached from a browser or the mobile app. Git traffic goes through a proxy so credentials stay out of the sandbox. You can watch the agent work, steer it as it goes, and run several tasks across different repositories at once without touching your own set-up.
What changes
The agent moves off individual laptops and onto a shared service. Instead of installing and updating Claude Code on every machine, the team signs in to the same place. Three things follow from that.
- Onboarding gets much quicker. A new developer needs a browser and a GitHub connection, and that's it. No fighting with PATH variables, clashing extensions or an out-of-date CLI.
- Security is handled in one place. Each session runs in a sandbox with limits on network and file access. Anthropic does the hardening, so you aren't relying on every developer's workstation being locked down, and the proxy keeps the agent to repositories you've authorised.
- Parallel work becomes practical. You can start several sessions across different repositories from one screen, with a bug fix in one, a refactor in another and tests in a third, each ending in a pull request with a summary.
Why it matters for small businesses: if you work with freelancers or an outside developer, cloud sessions make access simpler. Someone can work through a sandboxed agent without you handing over full repository access or worrying about secrets leaking from their laptop.
What it does to a team's day
Take an ordinary day before cloud sessions:
- One developer fixes a bug in the login code on their own machine, commits and pushes.
- A second pulls the change, runs the tests locally and finds a conflict in the API layer.
- They set up a session, share screens and debug it together, which is an hour neither of them spends building anything.
With cloud sessions it goes differently:
- The first developer describes the fix in the Claude Code web interface. The agent works in its sandbox, runs the tests and pushes a branch.
- The second sees the pull request in GitHub, reviews it and, if something needs a closer look, starts their own session on the API layer from the same browser.
- Whoever is leading the work can see both sessions and how far each has got.
Local terminals don't go away. Some work still needs direct access to a machine. But for routine fixes and running jobs side by side, there's less switching between tools and the team stays in step.
Security and oversight
The obvious worry with a cloud agent is where your code goes. Anthropic's answer has several layers:
- Isolated sandboxes. Each session starts in a fresh environment. Network access is restricted by default, and you allow specific domains, such as the npm registry, when a task needs them.
- Scoped credentials. GitHub access goes through a proxy limited to the repositories you've authorised, rather than giving the agent your tokens.
- A record of what happened. The agent's actions are visible in the session history, so you can check afterwards what it did and why.
For businesses in regulated sectors this can make compliance easier. Rather than securing every developer's workstation, you rely on the provider's security, in the same way you already rely on your hosting company.
What it means for your website project
If we're building or looking after your website, cloud agents change some of what happens behind the scenes, and a little of what you see:
- Routine jobs come back sooner. Bug fixes, content changes and small SEO jobs can run in parallel sessions, which shortens the gap between asking and seeing it live.
- You can see what was done. Each change comes with a summary and a pull request, so it's easy to follow what changed and why.
- Standards stay consistent. Every session runs in the same controlled environment, so the same checks and tests apply every time.
None of this takes people out of the loop. Our designers and developers still direct the agent, review what it produces and make the decisions. It just takes the repetitive work off their hands so they can spend more time on the parts of your site that need thought.
Trying it yourself
To use Claude Code cloud sessions you need:
- a Claude Pro, Max, Team or Enterprise plan with the right seats
- GitHub repositories you're happy for the agent to work in
- to visit claude.com/code and connect your first repository
Start with something small and well defined, like fixing a typo or adding a missing meta tag, to get a feel for how it works. Move on to bigger jobs once you trust the process.
The mobile app is handy for checking on a long-running session when you're away from your desk.
Cloud agents don't replace developers. What they do is make working together smoother, keep security in one place and make the work more predictable. For a small business that relies on outside help or a spread-out team, that means fewer delays and more time spent on a website that works for your customers.
Want to see how AI-powered development works for your website?
We use tools like Claude Code cloud sessions to build and maintain AI-managed websites that are fast, secure, and easy to update. Our process combines agent efficiency with human oversight to deliver sites that perform.
Explore AI Website Design