24 Claude Cowork on Web and Mobile: Hand Off Work Anywhere
Originally published July 7, 2026.
24.1 Work Without a Desk
The desktop version of Claude Code was built for deep work — long sessions, complex codebases, multiple parallel agents. It remains the best tool for that kind of focused, high-intensity development. But deep work isn’t the only kind of work.
Most knowledge workers spend significant parts of their day away from their desk: commuting, in meetings, between appointments, at home in the evening. The work doesn’t stop during those times — questions arise, decisions are needed, tasks need to be checked on. But without access to your agents, those moments are wasted.
Claude Cowork on Web and Mobile brings agent-powered work to every device. It’s not a replacement for the desktop experience; it’s a complement that lets work continue wherever you are.
24.2 Three Changes
The expansion of Cowork to web and mobile introduces three fundamental changes to how work flows.
24.2.1 1. Work Follows You
Every session, every task, every agent interaction is now synchronized across all your devices. Start a task on desktop, check on it from your phone during lunch, review the results on the web from a different computer.
Desktop (Morning)
└─ Started: "Refactor the authentication module"
Agent is working...
Mobile (Lunch)
└─ Notification: Agent has a question
"Should I keep the legacy token format for backward compat?"
You reply: "Yes, keep it for now."
Agent continues working...
Web (Afternoon, different machine)
└─ Check progress: Agent completed the refactor
Review the diff, approve the changesThe synchronization is real-time and bidirectional. Changes made on one device are instantly visible on all others. The session state — context, conversation history, tool results — is preserved across devices.
Cross-device sync means you never lose context. You can start a complex task at your desk, hand off instructions from your phone when the agent needs direction, and review the final results from any browser. The work is tied to you, not to a machine.
24.2.2 2. Work Continues in Background
Perhaps the most powerful change: agents continue running even when no device is online. Previously, if you closed your laptop or lost connectivity, agent sessions would pause or terminate. Now, scheduled tasks and background sessions run on Claude’s infrastructure independently of your devices.
This means:
- A code review you started before leaving the office completes while you’re in transit
- A routine that runs at 2 AM executes regardless of whether your laptop is open
- A long-running migration continues overnight even if your machine goes to sleep
- A background agent keeps working while you’re in a meeting
# This routine runs at 2 AM — no device needed
routine:
name: "nightly-test-suite"
schedule: "0 2 * * *"
prompt: "Run the full test suite and report any failures"
notify_on:
- failure
- completionWhen you pick up your phone the next morning, you’ll find a notification with the results. The work happened; you just weren’t there to watch it.
24.2.3 3. Decisions Still Come to You
Agents are more autonomous than ever, but they still need human judgment for important decisions. Phone notifications ensure that when an agent needs your input, you know about it — wherever you are.
Decision notifications are designed to be actionable from the notification itself:
┌─────────────────────────────────────────┐
│ 🔔 Claude Code │
│ │
│ auth-refactor needs your input: │
│ │
│ "I found a deprecated API call in │
│ the user service. Should I replace │
│ it with the new endpoint, or leave │
│ it for a follow-up PR?" │
│ │
│ [Replace now] [Leave for later] [Open]│
└─────────────────────────────────────────┘You can respond directly from the notification without opening the app. For more complex decisions, tapping “Open” takes you into the full session context.
Notification types include:
| Notification | When | Action |
|---|---|---|
| Needs decision | Agent needs your input | Reply inline |
| Task complete | Session finished successfully | Review results |
| Error encountered | Agent hit an unrecoverable error | Diagnose |
| Long-running | Session exceeded expected time | Check progress |
| Routine triggered | A scheduled routine ran | View summary |
24.3 The Mobile Experience
The mobile app is designed for the moments between deep work — not as a replacement for the desktop. It excels at:
24.3.1 Checking on Agents
See all your active and recent sessions at a glance. Each session shows its current status, last activity, and any alerts.
┌─────────────────────────────────┐
│ Your Agents │
├─────────────────────────────────┤
│ │
│ ● auth-refactor 12m │
│ Running tests... │
│ │
│ ● pr-review-1234 3m ⚠ │
│ Needs your input │
│ │
│ ○ deploy-staging Done ✓ │
│ All tests passed │
│ │
│ ○ docs-update Done ✓ │
│ 3 files updated │
│ │
└─────────────────────────────────┘24.3.2 Quick Interactions
Respond to agent questions, approve changes, and send quick instructions — all from your phone.
24.3.3 Starting New Tasks
Need to kick off a task while away from your desk? Start a new agent session from mobile:
"Review the latest PR from Sarah and leave
constructive feedback. Focus on the database
queries — make sure they're using indexes."The agent starts immediately on Claude’s infrastructure. By the time you’re back at your desk, the review is done.
24.3.4 Reviewing Results
View diffs, read summaries, and check outputs from completed sessions. The mobile diff viewer is optimized for quick scanning — you can approve or reject changes with a tap.
Use mobile for triage and decisions, not for deep work. If an agent’s output needs careful review, flag it and come back to it on desktop. The mobile experience is optimized for speed, not depth.
24.4 The Web Experience
The web version of Cowork brings the full Claude Code experience to any browser — no installation required. This is valuable for:
- Working from different machines (office, home, travel)
- Quick access when you don’t have the desktop app installed
- Sharing sessions with colleagues (via secure links)
- Reviewing work on a larger screen when mobile isn’t enough
The web experience includes:
- Full session management
- Code viewing and editing
- Diff review and approval
- Terminal output (read-only)
- Agent configuration
24.5 One Home for Chat and Cowork
With the web and mobile expansion, Chat and Cowork now share one home. Previously, Claude Code sessions lived in a separate interface from Claude chat conversations. Now, they’re unified.
┌─────────────────────────────────┐
│ Claude │
├─────────────────────────────────┤
│ │
│ RECENT │
│ │
│ 💬 Chat: "Explain transformers"│
│ 2 hours ago │
│ │
│ 🤖 Agent: auth-refactor │
│ Running · 12 min │
│ │
│ 🤖 Agent: pr-review-1234 │
│ Waiting for input · 3 min │
│ │
│ 💬 Chat: "Draft an email" │
│ Yesterday │
│ │
│ 🤖 Agent: nightly-tests │
│ Completed · 6 hours ago │
│ │
└─────────────────────────────────┘This unification reflects how people actually work. A question you ask in Chat might lead to a task you hand off to an Agent. An Agent’s results might prompt a follow-up conversation. Keeping everything in one place means the context flows naturally between asking, doing, and reviewing.
24.5.1 Starting from Chat
You can start an agent session directly from a chat conversation:
You: "Can you refactor the authentication module
to use the new token service?"
Claude: "I can help with that. This is a multi-step
task that would benefit from an agent session.
Want me to start one?"
You: "Yes"
Claude: → Starting agent session "auth-refactor"...
The agent will work on this and notify
you when it needs input or completes.The chat conversation provides context that the agent inherits, so you don’t have to re-explain what you just discussed.
24.6 When to Use What
| Surface | Best For |
|---|---|
| Desktop | Deep work, complex tasks, parallel agents, heavy code editing |
| Web | Medium tasks, working from different machines, session review |
| Mobile | Quick decisions, monitoring, starting tasks, triage |
Desktop remains the primary surface for deep work. The web and mobile experiences are designed to extend your reach, not replace the full development environment. If you find yourself doing complex code review on mobile, you’re probably better off waiting until you’re at a desk.
24.7 Privacy and Security
Cross-device sync raises natural privacy and security questions. Anthropic’s approach:
- End-to-end encryption: Session data is encrypted in transit and at rest
- No training on your data: Your code and conversations are never used to train models
- Device authentication: Each device must be authenticated to your account
- Remote session termination: You can revoke device access from any other device
- Audit log: All access is logged and reviewable
# View your connected devices
claude devices list
# Revoke a device's access
claude devices revoke --device-id <id>24.8 Real-World Workflow
Here’s what a typical day looks like with Cowork across devices:
7:00 AM (Phone, at home) - Wake up to notification: nightly test routine completed, 2 failures - Open the session on phone, read the failure summary - Start a new agent: “Investigate the 2 test failures from last night” - Agent begins working on Claude’s infrastructure
8:30 AM (Desktop, at office) - Open Claude Code desktop - The investigation agent has identified the root cause - Review the agent’s analysis, approve the fix - Start the auth refactor task you’ve been planning - Background it with /bg
12:00 PM (Phone, at lunch) - Notification: auth refactor agent has a question about backward compatibility - Reply from phone: “Keep the legacy format for now” - Agent continues working
2:00 PM (Desktop, at office) - Auth refactor is complete - Review the full diff, run additional tests in the integrated terminal - Merge the changes
6:00 PM (Phone, commuting) - Start a new agent: “Generate documentation for the auth refactor” - Agent begins working
9:00 PM (Web, at home laptop) - Documentation agent completed - Review the generated docs, make a few edits - Approve and merge
Work flowed seamlessly across devices, with no context lost and no time wasted waiting.
24.9 Team Collaboration
Cowork’s cross-device sync also enables new forms of team collaboration. Since sessions can be shared and work continues in the background, teams can hand off work seamlessly:
- Session sharing: Share a read-only link to a session so colleagues can review progress
- Follow-the-sun workflows: A developer in San Francisco hands off a session to a colleague in Singapore at the end of their day
- Async code review: An agent runs a review and posts results; the author responds from their phone when convenient
- Pair debugging: One engineer starts a debugging session, another joins from a different device
San Francisco (5 PM):
Engineer A starts: "Investigate the payment gateway timeout"
Backgrounds the session, leaves for the day
Singapore (9 AM, next day):
Engineer B picks up the session from Agent View
Reviews the investigation so far
Continues the debugging with fresh context
Posts findings to the team channelFor follow-the-sun workflows, use the session handoff pattern: end your session with a clear summary of what you’ve found and what the next steps are. The engineer picking up the session will inherit the full context plus your summary, making the handoff seamless.
24.10 Key Takeaways
- Cowork on web and mobile extends Claude Code beyond the desktop, letting work continue wherever you are.
- Work follows you across devices — sessions, tasks, and agent interactions are synchronized in real-time.
- Work continues in background — scheduled tasks and background sessions run on Claude’s infrastructure, even with no device online.
- Decisions come to you via phone notifications — actionable alerts let you respond to agent questions directly from the notification.
- Desktop remains the primary surface for deep work — mobile and web are complements, not replacements.
- Chat and Cowork now share one home, allowing natural flow between conversation and task delegation.
- End-to-end encryption, no training on your data, and device authentication protect your work across devices.
- The result is continuous workflow — start a task at your desk, direct it from your phone, review results on the web.