jira-delivery-mcp

Jira Delivery MCP is a Jira operations toolkit centered around a local MCP server, reusable policy modules, and an optional Codex skill package.

It is built for teams that want AI-assisted Jira execution without losing deterministic tooling, workflow awareness, or auditability.

Use this project if you want an assistant to help with Jira work such as:

• creating and updating issues safely
• linking blockers and dependencies explicitly
• discovering the real workflow of a target project before acting
• selecting the next eligible issue instead of guessing
• bootstrapping a new project from a brief
• layering Codex skills on top of a deterministic Jira tool surface

• run a local MCP server for Jira and optional Confluence operations
• create, update, transition, comment on, and link Jira issues
• inspect project workflow and derive lifecycle semantics from the real tenant
• bootstrap starter backlog structure from a project brief
• store execution metadata in issue descriptions and resolve it during execution
• validate dependency state before picking or starting work

Typical flow:

1. Start the MCP server.
2. Read a target project and discover its workflow.
3. Create or refine starter issues from a brief.
4. Link blockers with explicit Jira issue links.
5. Pick the next eligible issue.
6. Transition it using a Jira-provided transition, not a hardcoded assumption.

| Area | Status | Notes | | --- | --- | --- | | Jira Cloud | Supported | Current implementation targets Jira Cloud REST APIs | | Confluence integration | Optional | Only active when Confluence config is present | | Team-managed projects | Supported | Discovery and execution are tenant-aware | | Company-managed projects | Supported | Admin behavior is more sensitive and should be validated carefully | | Workflow mutation | Supported with caution | Validate before rollout; treat as high-impact admin work