Kepler Documentation

Kepler Quick Answers: Running Coding Agents

Last updated: November 2026

In Kepler, find a short answer to a common question about running coding agents such as Claude Code and Codex, then follow its link for the full details. Start here when you know what you want to do but not where to find it.


Run agents where you work

Kepler runs agents on your computer, in WSL, or on another machine.

Common questions about where and how your agents run:

What you want to do Short answer Full details
Run several agents at once without their changes colliding In Kepler, each task gets a worktree, a private Git working copy. Agents change files there, so several can run at once without touching each other’s work or your own checkout. Tasks and Resources → Worktrees
Run Kepler on Windows Kepler installs natively on Windows 10 and later, on 64-bit and ARM64. It also runs on macOS 12 and later, and on Linux. Getting Started → Install Kepler, Run Multiple Coding Agents in Parallel on Windows
Run agents in Windows Subsystem for Linux (WSL) In Kepler, a WSL 2 distro is a remote environment. Kepler finds your version 2 distros and lists them under Found on this machine in Settings. Remote Environments → WSL environments, Manage Claude Code and Codex Sessions Across Accounts and SSH Hosts
Run agents on another machine In Kepler, add the machine as an SSH host. Kepler installs itself on the host over SSH, so you do not need to install anything there first. Remote Environments → Add an SSH host, Manage Claude Code and Codex Sessions Across Accounts and SSH Hosts
Use several accounts of the same agent In Kepler, Claude Code, Codex, GitHub Copilot, Auggie, and Grok Build can each hold more than one signed-in account. To add one, open Settings → Agents, expand the agent, and click Add account. Agent Sessions → Several accounts for one agent, Manage Claude Code and Codex Sessions Across Accounts and SSH Hosts

Start work

You can start a task from an issue in your tracker or from nothing at all.

Common questions about starting work, giving agents your rules, preparing each new worktree, and running agents on their own:

What you want to do Short answer Full details
Start an agent from a Jira, Linear, or GitHub issue In Kepler, connect your tracker, then run an Action from the issue’s row, and the agent starts with the issue attached. For a Jira issue, or a Linear issue not synced with GitHub, add a repository in its panel first. Issue Tracker Integrations, Actions, Start a Coding Agent from a Jira or GitHub Issue and Take It to a Pull Request
Start without an issue In Kepler, click New task or press Ctrl/⌘ + N, then write a prompt. You can attach nothing at all and turn the task into real work later. Getting Started → Start something that is not tracked yet
Move my work over from Conductor In Kepler on macOS, click the import card on an empty Tasks list, or choose File → Import from Another App…. Each Conductor workspace becomes a task, and its chats continue in Kepler. Create a task → From another agent app, Move from Conductor to Kepler
Give every agent on a task your team’s rules In Kepler, attach a Note to the task with the rule, constraint, or decision. Every session on the task gets it, including sessions you start later. Tasks and Resources → Notes
Copy .env files into every new worktree In Kepler, list them in a .worktreeinclude file at the repository’s root, in .gitignore syntax. Kepler copies the matching Git-ignored files into each new worktree. Tasks and Resources → Make a new worktree ready to build, Set Up Every New Agent Worktree Automatically
Install dependencies in every new worktree In Kepler, add a per-repository command, such as pnpm install, and turn on Run on worktree creation. Settings → Per-repository commands, Set Up Every New Agent Worktree Automatically → Install dependencies
Keep agents in different worktrees from colliding on ports or databases In Kepler, a worktree separates files, not ports, databases, or caches. Use a command that runs on worktree creation to give each worktree its own settings, such as its own port or database name. Tasks and Resources → What a worktree does not separate, Set Up Every New Agent Worktree Automatically → Own port or database
Run an agent on a schedule, or when an issue or pull request changes In Kepler, click Automations on the Dashboard and create an automation. It runs an Action or your own prompt on a schedule, or when new GitHub or Jira activity matches its query. Automations

Keep track of your agents

Kepler shows you what your agents are doing and when they need you, on your computer or your phone.

Common questions about watching and answering your agents:

What you want to do Short answer Full details
See what every agent is doing In Kepler, the Agent Graph draws every task, session, turn, tool call, subagent, and file the agents touch. It updates live as they work. The Agent Graph, Keep Track of Several Claude Code and Codex Sessions
Find out what needs you In Kepler, open Ask Kepler and ask. Its chat can see what is on your screen and check on your tasks and sessions. Ask Kepler, Keep Track of Several Claude Code and Codex Sessions → See which session needs you
Answer an agent’s permission request In Kepler, the request pins to the top of the session. Choose Allow, Allow for this session, Allow always, Reject, or Reject always. Agent Sessions → Permission requests, Approve Coding Agent Requests from Your Phone → Answer a permission request
Check on your agents from your phone In Kepler, turn on Remote Access and scan its QR code with your phone’s camera. Kepler opens in your phone’s browser, and your work keeps running on your computer. Remote Environments → Using Kepler on a phone, Approve Coding Agent Requests from Your Phone

Review and ship

When an agent finishes, its work is not necessarily ready to ship.

Common questions about reviewing an agent’s work and getting it merged:

What you want to do Short answer Full details
Have the agent prepare its work for review In Kepler, create an Action that asks the agent to summarize what it changed, what it tested, and what it did not test. The example Prepare for review Action shows the prompt. Actions → Example: Prepare for review, Review an AI-Written Pull Request Before Your Teammates Do
Read what an agent changed In Kepler, open the task’s worktree and read the diff in the Changes overlay. From there you can comment, stage, and commit. Review What a Coding Agent Changed
Undo an agent’s change to a file In Kepler, use Discard changes, which asks first. Your current version goes to the Trash (the Recycle Bin on Windows), so you can get it back. Review What a Coding Agent Changed → Staging and committing
Have a different agent review the work In Kepler, set the Review Action to a different agent from the one that wrote the code. Actions → Review with a different agent, Review an AI-Written Pull Request Before Your Teammates Do
Apply the review comments on a pull request In Kepler, run Address Feedback on the pull request. The agent works through each comment, applies the changes it agrees with, pushes back on the rest, and replies to each comment. Actions → The four built-in Actions
Open a pull request In Kepler, you or the agent opens the pull request, and Kepler attaches it to the task once it exists. The example Open a pull request Action has the agent open it for you. Review What a Coding Agent Changed → Opening a pull request, Actions → Example: Open a pull request, Start a Coding Agent from a Jira or GitHub Issue and Take It to a Pull Request

Have feedback about this article? Did we miss something? Let us know!
On this page