Features

Kanbus ships a single binary that handles every project-management operation from your terminal. Create issues, update statuses, assign work, promote local tasks, and query your backlog—all without leaving your shell or opening a browser.

Every command returns structured output that coding agents can parse directly. Agents read requirements with kbs show, update progress with kbs update, and list work with kbs list. No MCP tools, no API tokens—just a CLI that works like any other Unix tool.

Every issue is a JSON file in your repository's project directory. When you create an issue, Kanbus writes a file. When you update one, it updates that file. There is no background daemon, no SQLite database, and no index to corrupt or rebuild. Your project data is always right there in the working tree, readable by any tool that can open a file.

Because issues live in Git, you get complete change history automatically. Branch your backlog alongside your code, review issue changes in pull requests, and merge project-management updates through the same workflow as everything else. An optional overlay cache keeps listing fast, but it is entirely disposable—delete it any time and Kanbus rebuilds from the files.

Pull Jira issues directly into your repository as local JSON files. Coding agents get full context—titles, descriptions, statuses, priorities—without making API calls or configuring MCP tools. The issues are just files on disk, available the moment the agent starts working.

Sync runs on demand via the CLI. Kanbus maps Jira fields to its own lightweight schema, so your agents work with a consistent format regardless of how your Jira project is configured. Updates flow one way: from Jira into your repo. Your local workflow stays independent.

View and manage issues from multiple repositories in a single workspace. Virtual projects let you define a collection of repos and see all their issues on one board, without moving files or changing directory structures.

This is useful for teams that split work across microservices or monorepo boundaries but still want a unified backlog. The CLI and console both support virtual project views, so you get the same filtering and status tracking you would in a single repo.

The full Kanbus Kanban board embedded in your editor as a VS Code extension. View issues, drag between columns, and update statuses without switching windows. The board renders in a webview panel alongside your code.

The plugin communicates with the same local Kanbus server used by the standalone console, so changes made in VS Code are immediately visible in the terminal and vice versa. When the gossip layer is active, updates from other agents appear in real time.

Define project rules and workflows using Gherkin BDD syntax, stored as .feature files in your repository. Policies can enforce hard constraints—like requiring a description before an issue moves to In Progress—or provide soft guidance that agents receive at the right moment.

Kairotic policies fire contextually: when an agent tries to perform an action, the policy engine evaluates the relevant rules and returns guidance inline. This keeps agents on track without front-loading a massive prompt. Policies are versioned in Git and reviewable like any other code.

A dedicated metrics panel in the Kanbus console that shows issue counts, status distributions, and type breakdowns. All metrics respect the current filter state, so you can drill down into a specific project, scope, or label and see accurate totals for just that slice.

Metrics update in real time when the gossip layer is active. The panel is designed for quick standups and progress checks—see at a glance how many issues are open, in progress, or blocked without running a separate query.