Kanjin for opencode
opencode, from a ticket to a pull request.
Kanjin runs opencode on your Mac, with the providers and models that you set up in it. It brings the ticket, a worktree in each repository, its own tools, and a diff that folds the mechanical edits. opencode writes the code, and you review it.
A window on opencode serve
Kanjin starts opencode serve on 127.0.0.1 with a random password. The Stream view shows the messages, the permission requests and every change since the session started. The Terminal view runs opencode attach on the same session.
Join a Session continues a session of an opencode server that already runs. At the end of each turn, the classifier folds the blocks that preserve behavior.
Jira tickets in opencode, without an MCP setup
Open a Jira ticket and ask opencode about it. The first message holds a brief: the description and the comments, the repositories and pull requests that name the ticket, and the pages that it links to.
Kanjin gives opencode its own MCP server for the session, with a new token each time. opencode reads the ticket and its pages, and drafts a comment or a status change for your approval.
Edits only in the worktrees
Until the ticket has worktrees, opencode runs with an agent that denies every tool except the read tools. In Implement and Fix, it may edit files only in the ticket’s worktrees.
opencode has no sandbox for commands, so Kanjin asks you before each command. An approved command can write outside the worktrees, and the start panel says so before Implement.
Review before the pull request
Review reads the changes of every worktree and lists defects, and Overengineer looks for needless layers. Either can run on a second agent, such as Codex, and report back to the opencode session.
Fix checks each finding against the code and the ticket, fixes the valid ones and skips the rest. It does not commit. Then you read the diff, with the mechanical edits folded.
Your providers, your tools
Kanjin builds no agent of its own. It runs the opencode that you installed, with its own providers and sign-ins.
In a ticket session, opencode also gets the web and your own MCP servers. These tools work with your accounts, and the start panel says so.
Sessions that stay, and chains that run
A ticket session stays when you open another item, and it continues on the same opencode session after Kanjin restarts.
A chain runs up to eight steps of a ticket in order. An automation starts a saved chain when a ticket enters a Jira list, and every approval still waits for you.
Questions
Is Kanjin a GUI for opencode?
In part. Kanjin shows opencode’s own terminal interface and a Stream view of the same session. It is also a development environment around it: Jira tickets, git worktrees, the steps of a ticket, and a review of each change.
Which models work?
The providers and models that you set up in opencode. Kanjin does not choose the model, and it holds no model keys.
Does Kanjin change my opencode configuration?
No. Kanjin passes its MCP server and the rules of a session in the environment variable OPENCODE_CONFIG_CONTENT. Your configuration files stay as they are.
Why does opencode ask before each command?
opencode has no sandbox for commands, so Kanjin cannot limit where a command writes. It asks you instead, and the start panel says so before Implement.
Is Kanjin made by the opencode team?
No. Kanjin is an independent product, not affiliated with opencode.