Run multiple AI agents
at once, from your phone.
You can run multiple AI agents at once by launching separate Claude Code, Codex, or Cursor sessions, often in git worktrees or tmux panes so they do not collide. The hard part is not starting them, it is watching them. Pushary sends a push from each agent when it finishes or needs a decision, and lets you approve from your phone, so one person can supervise a fleet.
Hacker News'te yer aldıHow to actually run several agents
Parallel agents work when each one has its own space. Keep their files, branches, and terminals separate so they never step on each other.
One git worktree per task
Give each agent its own worktree so two sessions never edit the same files or fight over the same branch. Same repo, isolated working directories.
Separate terminals or tmux panes
Launch a Claude Code, Codex, or Cursor session in each terminal or tmux pane. Split the window so every agent has its own scrollback.
Keep repos isolated
For unrelated projects, run agents in different clones or directories. Scope each one to a task so their output does not collide.
The supervision problem
Starting agents is the easy part. The cost shows up when you try to keep an eye on all of them at the same time.
You cannot watch N terminals
Starting five agents is easy. Watching five terminals is not. The moment you focus on one, the other four might finish or stall without you noticing.
Missed prompts stall the fleet
An agent that needs a decision sits idle until you happen to look at its pane. Context-switching between panes wastes the parallelism you set up.
Supervise the whole fleet from one phone
Pushary sits above your parallel agents. Each one reaches you on its own, and you decide from your phone.
A push from each agent
Every agent sends a notification the moment it finishes or hits a decision, so you do not have to sit on any single terminal.
Each agent identified
Notifications name the agent and session, so you know which one is asking before you even open your phone.
Fleet view in one place
See your parallel agents together and tell at a glance which one is blocked on you and which is still working.
Approve from your phone
Answer questions and approve risky steps from your lock screen, so one person can supervise a whole fleet from anywhere.
Run a fleet, supervise from your phone.
Every agent pings you when it finishes or needs a decision. Set up in two minutes.
Set up your agent
Frequently asked questions
How do I run multiple AI agents at once?
Launch a separate Claude Code, Codex, or Cursor session for each task, and give each one its own git worktree or directory so they do not edit the same files. Run each in its own terminal or tmux pane. Then connect Pushary so every agent pings you when it finishes or needs a decision, and you can supervise all of them from your phone instead of watching each terminal.
How do I run Claude Code overnight?
Start the session, set the task, and connect Pushary so the agent notifies you when it finishes or gets blocked on a decision. You approve from your phone, so the agent keeps moving while you are away from the terminal instead of stalling until morning.
How do I keep track of several agents at the same time?
You cannot reliably watch several terminals at once. Pushary sends a push from each agent when it finishes or needs input, names which agent is asking, and shows your parallel agents in one fleet view. That lets one person supervise a fleet without staring at every pane.