Create a Task

Last updated: June 2026

What is a Task?

A Task is the core unit of work in Kepler, GitKraken’s Agentic Development Environment (ADE). A Task holds work across one or more repos.

Every Task contains:

• One or more worktrees — one Git worktree per repo
• One or more agent sessions — running coding agents within the Task
• A diff and changes view
• Shared context — standing instructions sent to every agent session in the Task

---

Three ways to create a Task

Kepler gives you three starting points depending on where your work lives:

Starting point	When to use
From scratch	You know the goal and don’t need to pull context from an existing issue or PR.
From an issue	You want Kepler to pull issue title, description, and metadata into the agent’s context automatically.
From a pull request	You want an agent to start a review or address open review comments on an existing PR.

---

Create a Task from scratch

Start here when you have a clear goal and no existing issue or PR to pull context from.

1. Click + New task.
2. Select one or more repos.
3. Choose New worktree or select an existing branch.
4. Set the base branch using the From dropdown. The dropdown is searchable; origin/main is labeled DEFAULT.
5. Optionally click + Add repo to include additional repos in the same Task. Each repo gets its own worktree and base branch configuration.
6. Set Agent, Mode, Model, and Effort in the bottom bar.
7. Click Launch task.

Settings reference

Setting	What it controls	Default	Options
Repo(s)	Which repos are included in the Task	—	Any connected repo
Worktree	Whether to create a new Git worktree or use an existing branch	New worktree	New worktree, existing branch
Base branch	The branch the worktree is created from	origin/main	Any branch (searchable)
Agent	Which coding agent runs sessions in the Task	—	Configured agents
Mode	Agent operating mode	—	Depends on agent
Model	The underlying language model the agent uses	—	Available models
Effort	How much work the agent does before pausing for review	—	Low, Medium, High

---

Create a Task from an issue

Use this path to have Kepler pass issue context (title, description, and metadata) to the agent automatically.

Click + New task, then select the Start from issues tab.

Launching Tasks from issues

1. Select one or more issues.
2. If you selected multiple issues, choose how to group them:
   • Split into N tasks — creates one Task per issue. Use this when each issue represents independent work.
   • Group as 1 task — places all issues under a single Task. Use this when the issues are related and you want one agent to work across them.
3. In the Task preview row, confirm or change the repo, worktree type, and base branch for each Task.
4. Click + Add repo on any Task row to span that Task across multiple repos.
5. Set Agent, Mode, Model, and Effort in the bottom bar. These settings apply to all Tasks being launched.
6. Review the summary (e.g., “3 issues → 3 tasks”), then click Launch task.

Kepler creates one Task per selected issue and passes the issue title, description, and metadata to the agent when the session starts.

When multiple issues are selected, the Split into N tasks and Group as 1 task buttons appear above the Task preview rows. Use the bottom bar to configure the agent, model, mode, and effort for all Tasks at once before launching.

Supported issue trackers

Kepler can pull issues from the following trackers:

• GitHub Issues
• GitHub Enterprise
• GitLab
• GitLab Self-Hosted
• Jira
• Linear
• Trello
• Azure DevOps

For integration setup, see Issue Tracker Integrations.

Finding an issue

Type in the search bar to filter issues by title, or paste an issue URL directly to jump straight to a specific issue.

Use the filter bar to narrow results further:

• Assigned to me / All visible toggle
• Provider — filter by issue tracker
• Repo — filter by repository
• Show tasked — shows issues that already have an active Task (hidden by default)

---

Create a Task from a pull request

Use this when starting a PR review or having an agent address open review comments.

Launching Tasks from a single PR

1. Click + New task.

2. Click Start from pull requests and select a PR from your connected providers, or paste a URL directly.

3. Before you launch, pick a mode:
   • Review — the agent analyzes the changes and suggests improvements.
   • Address feedback — the agent reads the review comments and applies the ones that make sense.

4. In the Task preview, confirm the repo and the worktree branch Kepler will create (e.g., “Worktree will be created on new-example-page“).
5. Set Agent, Model, and Effort in the bottom bar.
6. Click Launch task.

Launching Tasks from multiple PRs

1. Click + New task, then click Start from pull requests.
2. Select two or more PRs. Two options appear:
   • Split into N tasks — creates one Task per PR. Use this when each PR represents independent work.
   • Group as 1 task — places all PRs under a single Task. Use this when the PRs are related and you want a single agent to work across them.
3. Choose your grouping option.
4. Pick a mode (Review or Address feedback) for each Task row.
5. Confirm the per-task repo selector in each Task preview row.
6. Set Agent, Model, and Effort in the bottom bar. The summary shows “N PRs → N tasks.”
7. Click Launch task.

Supported PR providers

GitHub, GitHub Enterprise, GitLab, GitLab Self-Managed, Bitbucket, Azure DevOps.

For integration setup, see Pull Request Integrations.

Finding a pull request

Use the search bar to find PRs by title or by pasting a URL. Use the filter bar to narrow results by provider, repo, or Hide tasked.

---

Task detail: worktrees, sessions, and shared context

Once a Task is open, the Task header shows the Task name and the total worktree count (e.g., “5 worktrees”).

Worktrees

The Worktrees section lists every Git worktree in the Task with its branch name and repo.

Sessions

The Sessions panel lists all agent sessions in the Task. Each entry shows the branch, repo, and a NEW CONTEXT badge when there is unread context.

Shared context

Shared context is content you add to a Task that every agent session receives. Use it for standing instructions, style rules, or reference material that all agents should follow.

Example: “If a .md file has a date stamp, change the date stamp to June 2026.” Adding this to shared context means every agent session in the Task follows the rule without you repeating it per session.

To manage shared context:

• Each piece of shared context appears as a prompt card showing the prompt text and version number (e.g., “v2”).
• Click Edit on a card to update it or Remove to delete it.
• Click + Add markdown to add a new context block.

---

Manage Tasks

Session status vs. Task stage

Task stage is the Kanban column the Task occupies. It progresses through Exploration, In Development, In Review, and Done. Kepler advances the stage automatically based on agent activity — you cannot move it manually. Session status is the real-time state of an individual agent session.

Status	Meaning
🟠 Needs Attention	The agent is waiting for your input.
🟢 Active	The agent is running.
⚫ Idle	The session stopped but is not complete.
🔴 Errored	The session hit an error.
⚫ Inactive	The session is not running and has no pending work.
Disconnected	The session lost connection to the agent runtime.

Notifications

Kepler sends a toast notification when a Task completes or needs attention. The notification shows the Task name, the agent that ran it, and the repo and branch. Click View to jump directly to the Task.

Archiving a Task

Click the archive icon on the Task card to archive it. Archived Tasks are removed from the active board.

Delete a Task

Click the trash icon on the Task row to delete it. A Delete confirmation tooltip appears before the action completes.

---