Anthropic Claude Code on the Web Runs Async Agent Pull Requests

What changed

Anthropic opened Claude Code on the web in October 2025 for Pro, Max, Team, and allowed Enterprise seats. You describe a coding task in the browser at claude.ai/code. Anthropic starts a hosted Claude Code session on its cloud. The agent clones your linked GitHub repository, edits files, runs commands in a container, and can open a pull request when it finishes.

The product follows the same async agent pattern as OpenAI Codex Cloud and Google Gemini Jules. The developer machine can sleep while work continues. Session state persists if you close the tab, which matters for long refactors or test suites.

Official docs state the preview runs on Anthropic managed cloud setup. GitHub linkage uses the same login link app patterns as other coding agents. Teams should read the access scopes before they connect org wide.

Why it matters

Platform and security leaders now govern two Claude Code surfaces. The CLI runs on a laptop with local secrets and shell history. The web product runs in Anthropic cloud but still writes to production repos.

That split changes audit questions. Who approved repo access? Are agent branches protected? Do auto generated pull requests require human review? Rules teams that already restrict self hosted runners should apply similar rules here.

Product teams gain speed for chores such as test fixes, dependency bumps, and internal tools. The tradeoff is broader machine driven write access to source control if policies lag.

Who is affected

Engineering managers and developer experience owners choose whether to enable web sessions. Application security reviewers must map login link scopes and branch protections. GitHub organization admins inherit new third party app risk.

Individual contributors on paid Claude plans may start using web agents before IT publishes guidance. That shadow use is the practical trigger for policy work this quarter.

What to do next

Run a pilot on one non production repository. Require draft pull requests only and block auto merge. Document allowed task types such as docs and tests before you allow feature work.

Add a short security review checklist covering secret handling, dependency changes, and license rules on agent commits. Mirror controls you use for other async coding products.

What to watch

Track anthropics/claude-code issues for branch selection, enterprise admin toggles, and MCP connector support on web sessions. Any MCP expansion increases blast radius because tools can reach external APIs.

Watch competitor pricing and seat bundling. Async agents are becoming a default part of paid AI plans, which shifts budget from point CI add ons to model subscriptions.
### Plain terms recap

This desk note covers a real public update from the last few weeks. We read the primary link first. We drop claims that sources do not support. Your next step should match your team size and risk level.

Save the source links in your team decision log. Pin the versions or configs named above. Map the change to access rules you already use for code, cloud, or lab systems.

If the update touches user data, run a quick review with privacy peers. If it touches money, health, or safety systems, test in staging before you change production.

Fast vendor releases often outpace internal policy. That gap feeds shadow use. Publish a one page guide so teams do not guess about allowed tools.

Watch the primary link for patches or scope changes. We will note major fixes in later desks if facts change.

Good briefs name who must act this week. They also name what signal would prove the story wrong or outdated.

Share this brief with your legal, finance, or ops peers when the story touches their systems. One short email beats a long slide deck for early alignment.

If you already run a weekly risk review, add this topic for five minutes. Ask what would break if you did nothing for thirty days.

When in doubt, run a one hour spike with two engineers and one security partner. Write down what you learned before you scale access.

Print the source titles in your ticket system so future you can trace why a decision was made. Future teammates should not rely on chat memory alone.

If a vendor fixes a bug, rerun your smoke test the same day. Small patches still change behavior in agent and cloud tools.

Keep a single shared doc with dates, owners, and links for each desk you follow. That doc saves time when audits or board questions arrive.

Photo: cottonbro studio / Pexels

Sources

• Primary — Anthropic, Claude Code on the web (October 2025). Defines hosted sessions, GitHub linkage, and plan eligibility.
• Secondary — Simon Willison, October 2025 newsletter (October 2025). Independent description of async PR workflow.