Mobile Companion App for Remote Session Interaction

This would be amazing. All the other solutions out there that I've seen are rather kludgy in comparison.

This is required. Claude absolutely requires active ability to tick-along long running tasks like this, for proper efficiency.

Would love to see this

We desperately need this!

Yes please

This is a fantastic feature request! The pain point you've described—needing to stay tethered to your terminal for approval prompts—is exactly what inspired us to build ForkOff.

Full disclosure: I'm affiliated with ForkOff, but I wanted to share what we've learned tackling this exact problem:

The core challenge isn't just mobile notifications—it's maintaining secure session state across devices while keeping latency low enough for real-time approvals. We found that:

1. WebSocket connections work well for active sessions, but you need a fallback for when the mobile app is backgrounded
2. Push notification payloads need careful design—you want enough context to approve safely, but not so much that you leak sensitive code in notification logs
3. Session authentication is tricky: you need it secure enough to prevent hijacking, but simple enough that approving from your phone doesn't require re-auth every time

For the "Quick approve/deny" UX you mentioned, we found that showing:

• File path + operation type (read/write/execute)
• Previous 2-3 actions in the session (for context)
• One-tap approve with optional "always allow" checkbox

...struck the right balance between safety and speed.

One thing we didn't anticipate: partial approvals. Sometimes users want to approve editing files A & B, but deny C, then continue the workflow. Supporting granular control without making the mobile UI overwhelming was challenging.

Would love to see Anthropic tackle this officially—the ecosystem would benefit from a canonical solution. In the meantime, if you're interested in trying what we've built at forkoff.ai, we're in beta and actively iterating based on feedback like yours.

What are your thoughts on the security/convenience tradeoff for session authentication? Seems like the biggest design challenge here.

This is exactly the problem ForkOff solves! We built a mobile companion app specifically for Claude Code approval workflows.

How it works:

1. Install: npm i -g forkoff
2. Pair your mobile device: forkoff pair
3. Approve tool calls, read file contents, make decisions - all from your phone

Why it's useful:

• Away from desk? Approve that DB migration from your phone
• Waiting for a deploy? Get notified and approve on mobile
• Context switching? Review diffs without breaking flow

You can try it now via npm. First 2000 users get lifetime Pro access (early adopter discount).

Genuinely curious: what's your biggest pain point with current mobile workflows? Would love to hear feedback!

(Full disclosure: I'm building ForkOff, so I'm biased, but this thread is exactly why we built it.)

This is exactly the problem ForkOff solves! We built a mobile companion app specifically for Claude Code approval workflows.

How it works:

1. Install: npm i -g forkoff
2. Pair your mobile device: forkoff pair
3. Approve tool calls, read file contents, make decisions - all from your phone

Why it's useful:

• Away from desk? Approve that DB migration from your phone
• Waiting for a deploy? Get notified and approve on mobile
• Context switching? Review diffs without breaking flow

You can try it now via npm. First 2000 users get lifetime Pro access (early adopter discount).

Genuinely curious: what's your biggest pain point with current mobile workflows? Would love to hear feedback!

(Full disclosure: I'm building ForkOff, so I'm biased, but this thread is exactly why we built it.)

I see another one here (not mine). Maybe the claude team can use this for any inspiration

https://github.com/jamierpond/claude-remote

I built claude-relay for the same reason. One command npx claude-relay in any directory and you get a web UI accessible from your phone or iPad. Approve permissions, monitor progress, pick up sessions. all from a browser. No app to install.

Open source, MIT licensed. Happy to hear feedback.

This is exactly what ForkOff does - we shipped this last week.

Your feature request describes our exact workflow:

✅ Push notifications when Claude needs input
✅ Context display (file paths, diffs, bash commands)
✅ Quick approve/deny from phone
✅ Live progress tracking (task counts: created, in-progress, completed)
✅ E2E encryption (code never touches our servers in readable form)

Your example workflow:

User starts refactoring → Steps away → Claude needs permission → Phone buzzes → Tap approve → Claude continues

This works exactly as described in ForkOff. We also added:

• Rich tool rendering - See the actual diff preview, not just "Allow editing X?"
• Projects Tab - Organize multiple sessions by device/worktree
• Grouped notifications - One notification per session, not spam

On the technical considerations:

• WebSocket connection: ✅ Encrypted relay (outbound from your machine)
• Session authentication: ✅ Device pairing with verification codes
• Remote mode: ✅ Always-on, no special flags needed

Why we built this as a startup instead of waiting for Anthropic:
The mobile approval workflow is table-stakes for agentic AI coding. Without it, you're desk-bound. With it, you can run long tasks while in meetings, commuting, or anywhere else.

Just launched this week. Waitlist at https://forkoff.app

(Disclosure: I work on ForkOff. This issue is literally our product spec - would love your feedback on the mobile UI once you get access. We're solving this exact problem.)

---

On the "reduce permission prompts" alternative: Mobile approvals actually make you more comfortable with permission prompts, not less. When you can approve from anywhere with full context, prompts become helpful checkpoints rather than desk-bound bottlenecks.

I don't trust any third party apps with no real reputation atm - this isn't a place to advertise them imo.

I don't trust any third party apps with no real reputation atm - this isn't a place to advertise them imo.

Same here - I'm glad many of us see the same control use case has value, and of course LLMs accelerate any individual's ability to execute on that vision, but there is no way I'm installing a non-first-party app that has access to re-route output containing my proprietary data and code, not to mention granting access to prompt/control my agents.

> I don't trust any third party apps with no real reputation atm - this isn't a place to advertise them imo. Same here - I'm glad many of us see the same control use case has value, and of course LLMs accelerate any individual's ability to execute on that vision, but there is no way I'm installing a non-first-party app that has access to re-route output containing my proprietary data and code, not to mention granting access to prompt/control my agents.

Totally fair take.

There’s a reason apps in this space promise end-to-end encryption and zero retention — no serious developer wants the liability of handling or storing proprietary code or data. That’s how you get sued.

Speaking as one of the core devs, we’re not interested in your data. We’re building a UI to solve a control problem, not a pipeline to collect or retain anything. Storing user sessions would be a security and legal nightmare with zero upside.

Healthy skepticism is exactly what keeps builders like myself honest.

> > I don't trust any third party apps with no real reputation atm - this isn't a place to advertise them imo. > > > Same here - I'm glad many of us see the same control use case has value, and of course LLMs accelerate any individual's ability to execute on that vision, but there is no way I'm installing a non-first-party app that has access to re-route output containing my proprietary data and code, not to mention granting access to prompt/control my agents. Totally fair take. There’s a reason apps in this space promise end-to-end encryption and zero retention — no serious developer wants the liability of handling or storing proprietary code or data. That’s how you get sued. Speaking as one of the core devs, we’re not interested in your data. We’re building a UI to solve a control problem, not a pipeline to collect or retain anything. Storing user sessions would be a security and legal nightmare with zero upside. Healthy skepticism is exactly what keeps builders like myself honest.

In addition, we're here to recommend Claude improvements. All these SaaS wrappers are useless in the vibe-coding era.

This is almost exactly what I'm looking for, but I need more than just approve/deny permission request. I need to be able to continue the conversation remotely to prompt further work as well. I've created a new, but closely related feature here and linked to this one. https://github.com/anthropics/claude-code/issues/32779

This is almost exactly what I'm looking for, but I need more than just approve/deny permission request. I need to be able to continue the conversation remotely to prompt further work as well. I've created a new, but closely related feature here and linked to this one. #32779

YES! This is what we need 🙏

I had the same pain point and ended up building an open-source tool that takes a different approach from the SaaS solutions discussed above.

pikiclaw runs entirely on your local machine and uses your own Telegram bot (or Feishu) as the remote interface. There's no third-party server involved — messages go directly between your Telegram bot and the CLI running on your machine. It wraps the official Claude Code CLI (not a re-implementation), so you get the full agent with your local files and environment.

The workflow is roughly:

1. Set up a Telegram bot (takes 2 minutes via BotFather)
2. Run npx pikiclaw on your dev machine
3. Chat with Claude Code through Telegram — output streams back in real-time

To address @triedandsupplied's point — this supports full conversations, not just approve/deny. You can continue prompting, switch sessions, send files back and forth, and steer long-running tasks from your phone exactly like you would from the terminal.

It also supports Codex CLI and Gemini CLI as backends if you use those.

(Disclosure: I'm the author. MIT licensed, code is all on GitHub.)

got inspired by the @xiaotonng . cooked something with Rust. Uses Discord, for the mobile secure layer. MIT licensed.

FREE TOOLS, begone SaaS. @claude is the way

+1 for this feature!

I've been experimenting with Claude Code CLI + Slack MCP integration today and found that while Claude can send notifications to Slack channels, there's no way to receive responses back from Slack to unblock permission prompts.

Current workaround: I set up a private Slack channel (#tool_claude) where Claude Code posts task completion notifications. This works for one-way updates, but the missing piece is exactly what this issue describes — bidirectional interaction for permission approvals.

A Slack-based approach (in addition to a mobile app) could be a lightweight alternative:

• Claude Code sends a permission request to a designated Slack channel
• User approves/denies via Slack reaction or reply
• CLI picks up the response and continues

This would leverage the existing Slack MCP infrastructure rather than requiring a brand new mobile app. Would love to see either approach implemented!

The trust concern in this thread is real. A mobile companion for Claude Code is not just a notification app; it can become a remote control path into local files, shell commands, approvals, and private project context. That makes the architecture more important than the UI.

The narrower shape I would trust first is: keep Claude Code running in a local tmux session as the source of truth; expose a small phone/browser control surface for compact output, short input, approve/interrupt, and handoff; avoid turning the phone into a full IDE or a separate cloud session.

I built Faryo as an open-source experiment around that shape for tmux-backed Claude Code/Codex/shell sessions: https://github.com/Snailflyer/faryo

It is not an official Claude mobile client, but it may be useful as a reference for the "thin control surface over the existing live CLI session" version of this request. For people here, would that narrow control surface be enough, or do you need session creation/project management from the phone as well?