How to see what all your coding agents are doing at once
3 min read
One coding agent is easy to follow. You watch the terminal, answer when it asks something, and read the result. The trouble starts with the second and third. Now one agent is refactoring in one project, another is running tests in a second, and a third has been waiting on a yes-or-no question for ten minutes while you were looking somewhere else.
This post is about that problem: how to keep several agents in view without staring at each terminal in turn.
Why parallel agents get lost
Agents are slow in a way that invites multitasking. A task takes a few minutes, so you start another one. But each terminal looks the same from a distance: a block of scrolling text. You cannot tell from a glance whether the agent is still working, finished, or blocked on you.
The cost is not the waiting itself. It is the checking. Every time you cycle through windows to find the one that needs you, you lose your place in whatever you were doing.
What you actually need to know
For each agent, only three things matter at a glance:
- Working: leave it alone.
- Needs input: it asked a question or wants permission, and nothing will happen until you answer.
- Done: there is a result to read.
Everything else, the scrolling output and the intermediate steps, you only need when you decide to look closer. A good setup shows the three states for every agent in one place and lets you jump straight to the one that matters.
Keep terminals, browser and status together
Tidepane is built around this idea. Each terminal that runs a coding agent shows its live status in the sidebar, so you can glance across projects and see which agents are working, which need input and which are done. When one is waiting for you, you jump to it from the sidebar instead of hunting through windows.
Because the same window also holds a built-in browser, the pages an agent is working on can sit next to its terminal. If an agent is building a web app, your dev server, the docs and the running app are all in one workspace beside the terminal that started them.
A routine that works
A few habits make parallel agents manageable, whatever tools you use:
- Give each agent its own project or branch. Two agents editing the same files will step on each other.
- Start the slow tasks first. Kick off the long-running work, then do the quick things while it runs.
- Check status, not output. Look for "needs input" first. Read the full output only for agents that are done.
- Answer blocked agents quickly. A blocked agent is idle time you are paying for in attention.
- Review before you merge. Status tells you an agent finished, not that the result is right. Read the diff.
When to use a chat instead
Not every task needs a terminal. For work you want to hand off and check later, Tidepane also offers chats that keep working while you do something else, grouped by project. You can switch between chats and pages without interrupting them, and send a message while one works to steer it.
Start small
You do not need to run five agents on day one. Try two on separate projects, watch how the status helps you decide where to look, and add more once checking them feels effortless. The goal is not more agents. It is less time spent wondering what they are doing.