Your agents follow you, not the other way around
The most capable agent setup is useless during the hours you are not sitting at the machine that runs it. MuxAgent's mobile app turns agent supervision into something that follows you — review plans on the train, approve checkpoints during a break, monitor execution from wherever you are.
How mobile control works
Real-time event streaming
Agent events flow through a direct pipeline: the AI agent emits output, the ACP (Agent Communication Protocol) layer captures it as typed notifications — message chunks, tool calls, status updates — and the CLI daemon encrypts and forwards each event through the WebSocket relay to your phone. The mobile app renders these events in real time as a scrollable session view with a sticky input bar at the bottom. There are no polling intervals and no refresh buttons — output appears as it is generated, just as it would in a terminal, but on your phone.
Approval gates are modals, not warnings
When an agent reaches an approval checkpoint — after planning, before implementation, or at any custom gate in your workflow graph — the mobile app presents it as a high-contrast modal dialog. This is a deliberate design choice: approval is not a banner you scroll past. It is a blocking decision that occupies the screen until you act. The modal shows the agent's plan or the specific action requiring approval, and you either approve to continue or reject to redirect the agent. The agent waits at the checkpoint with no timeout, so you can start a task before a meeting and approve the plan whenever you are ready.
Content blocks for rich output
Agent output is not just text. The mobile app renders typed content blocks: text with syntax-aware formatting, images (screenshots, diagrams, generated charts), and structured metadata like tool call names and file paths. When an agent generates a screenshot during verification or produces a diff during implementation, you see the actual content on your phone, not a truncated text dump. This matters because meaningful approval requires seeing what the agent actually produced, not a summary of it.
Session management across machines
The mobile app provides a unified view of all your paired machines and their agent sessions. Each machine shows its hostname, connection status, and active sessions. You can start new tasks on any machine, switch between active sessions, and review completed ones — all from a single screen. Session metadata (ID, working directory, title, status, cost) is persisted in the app's local SQLite store, so your session history is available even when a machine is offline.
On-the-go workflows
Start before you leave, approve from anywhere
Kick off a task at your desk using the default workflow (plan → review → approve → implement → verify). Close the laptop and leave. The agent completes the planning stage and hits the approval checkpoint. Your phone shows the plan. You review it on the subway, approve, and the agent proceeds to implementation on the machine you left behind. When you arrive at your destination, implementation is done and the verification results are waiting.
Monitor long-running tasks during meetings
Some agent tasks take thirty minutes or more — large refactors, full test suite runs, multi-step deployments. Rather than blocking your schedule around the terminal, let the agent run and glance at progress on your phone. The real-time streaming means you see exactly where the agent is: mid-implementation, running tests, or waiting for your input. If something goes wrong, you know immediately instead of discovering it an hour later when you return to the terminal.
Triage and delegate from the couch
You notice a bug report in the evening. Rather than opening the laptop, you pull up MuxAgent on your phone, start a diagnostic task on the relevant machine using the plan-only workflow, and read the agent's analysis. If the fix is straightforward, you switch to the default workflow and let the agent implement it. If it needs more context, you have the agent's analysis ready for tomorrow. The phone is the natural triage interface because it is the device you already have in your hand.
Related reading
- Why managing AI agents from your phone changes everything
The phone is not a toy interface for coding agents. It is the natural control plane for work that outlives any single terminal session.
- Approval checkpoints are a feature, not a slowdown
The right checkpoint improves throughput by catching ambiguity early and focusing human attention where it matters most.
- Five workflow patterns for real-world agent tasks
Concrete recipes for matching workflow shape to task type — from quick fixes to multi-machine deployments.
- Remote control is the missing layer for AI coding agents
Why the current generation of coding agents needs a remote supervision layer to become a real developer tool.
Try it on your commute
Start an agent task before you leave the office. By the time you reach the station, the plan will be waiting for your approval on your phone. See the getting started guide to pair your first machine and experience mobile agent control firsthand.