Using BMAD-METHOD collaboratively in a monorepo + Jira integration (no double-entry)

[newBie974]

Hi BMAD team 👋

I really like this BMAD-METHOD ! I use it on personal project and feel comfortable with it.

We’re adopting BMAD-METHOD in a team setting (not solo) and we work in a monorepo (multiple apps/libs). We also need to keep Jira as the “source of truth” for management dashboards (sprints/velocity/burndown), while using BMAD for deeper specs, agent workflows, and implementation context.

I found the discussion about syncing with external PM tools (Jira/Trello/Linear) in #666, which is close to what we need.

Our context

• Monorepo: multiple packages/apps (shared libs + backend + frontend)
• We want BMAD artifacts (PRD/specs/stories/architecture) versioned in Git alongside code
• Jira stays the reporting layer for PM/management

What we’re trying to achieve (team workflow)

1. Collaborate with BMAD outputs without everyone generating conflicting docs
2. A repeatable structure for BMAD artifacts inside a monorepo (per app? per domain? per epic?)
3. Link BMAD story files ⇄ Jira issues (and ideally reduce “copy/paste double-entry”)
4. Clear “definition of done”: when do we update BMAD docs vs Jira fields/status?

Questions

1. Recommended repo structure for monorepos?

• e.g. /docs/bmad/... vs /apps//bmad/... vs /bmad/... at root
• do you recommend one “global PRD” + per-app story packs, or fully per-app?

2. Collaboration model: how do teams avoid conflicts?

• Do you treat BMAD outputs as “source docs” reviewed via PRs?
• Any tips for “one writer, many reviewers” vs everyone running agents?

3. Jira integration: what’s the current best approach?

• Is there an official or “blessed” integration pattern (MCP, scripts, templates)?
• If not official: what do you recommend pragmatically?

4. Mapping strategy (if doing sync): what fields should be mapped?

• Epic/Story key in BMAD frontmatter?
• Jira status ↔ BMAD status?
• Acceptance criteria / DoD: should we sync both ways or only one direction?

5. Best practice for workflows

• Example: run BMAD to produce PRD + stories → create Jira issues → dev work → PR links Jira → update story status

If you have a reference example repo or a canonical “team + Jira” setup, that would help a lot 🙏

[newBie974]

Gentle up 🙏 if someone could help us with this topics

[jrevillard]

Hey 👋

For question 3 (Jira integration), I built a BMAD extension module that does exactly this kind of sync — currently for GitLab Issues and GitHub Issues:

bmad-issue-tracking

It integrates directly into BMM workflows via TOML customization hooks (on_complete in sprint-status, dev-story, etc.) so that when you complete a story or update sprint status, the corresponding issue is automatically created/updated in your GitLab or GitHub project.

Key features:

• Syncs BMAD stories ⇄ GitLab/GitHub Issues (titles, descriptions, labels, status)
• TOML-based configuration (no scripts to maintain)
• Works as a --custom-source module via bmad install
• Respects branch patterns and project structure

Jira support is on the roadmap — the architecture is designed to be platform-agnostic, so adding Jira as a third platform is straightforward. If there's interest, I'd be happy to prioritize it.

For your question about field mapping (question 4), the module uses BMAD story frontmatter keys ({prd_key}, {story_key}, {epic_num}) and maps them to issue labels/milestones. Status sync is one-directional (BMAD → issue tracker) to keep things simple.

Would love feedback if you give it a try — especially to understand what Jira-specific fields you'd need mapped.

[JohnSmall]

I worked out a system using Claude with Jira and Confluence. After I'd worked it out I found out about BMAD. Much the same roles and responsibilities, except I'm doing it purely in Jira and Confluence instead of markdown files.

It's easy enough to implement the ideas in BMAD without the actual BMAD code, just plain Confluence + Jira + Claude. I found OpenAI/Codex needs a lot of handholding and constant attention. Claude seems to have enough common sense to work on its own.

Claude.ai, Claude Code with Atlassian integration is a brilliant combination. All the problems people have with Confluence and Jira, the time it takes to write tickets, the detail required to write good tickets, the chores etc. All go away with AI and you can use Confluence and Jira as it should be used. It becomes a source of truth with detailed specs linked to detailed tickets. I can trace all the way from PRD in Confluence to sprints to tickets, to commits and tags in the code.