XenJira reads the ticket, writes the change, tests it against the acceptance criteria and edge cases, scans it for security issues, and waits for your approval before anything lands.
Every run is transparent: you see the predicted cost before it starts, every agent step while it runs, and the real cost when it finishes.
Exact input-token count before each run, a low / expected / high price, and the real per-agent cost afterwards.
Prompt caching, Sonnet for code and Haiku for checks, targeted edits instead of full-file rewrites, and relevance-ranked context.
Anthropic, OpenAI, Google Gemini, Groq or DeepSeek. XenJira lists the models your key can use automatically.
Diff preview, PM review panel, protected branches, and a full runtime log. Nothing merges without approval.
Jira and AI keys live in your OS keychain and are used only for the duration of a run. Secret files are never sent as context.
Ticket-defined and generated edge cases are checked on every fix, with a coverage verdict in the review.
Try everything free, then one flat price. You pay your AI provider directly for tokens — no markup.