From idea to issues

Start with a user problem, agree on the scope, then create work your team can deliver.

The manual path: every plan

  1. Capture the idea. Create a PRD, explain the problem and desired outcome, and name the non-goals.
  2. Make scope testable. Add keyed requirements such as R1: a returning shopper can reuse approved checkout details. Set priorities, and link teams/projects.
  3. Review the scope. Submit the PRD for review. Have its owner or a workspace owner/admin approve the agreed version.
  4. Describe the user outcome. Add a story for R1 and write acceptance criteria. Split it if the outcome is too large to deliver and verify together.
  5. Review engineering tasks. Choose Create issues, add the tasks needed to satisfy the criteria, check team/project/points, then confirm.
  6. Follow delivery back to scope. Open the linked issues and the PRD Stories tab. Coverage shows requirements with stories; issue progress shows delivery. Reopen PRD review when substantive scope changes.

The AI-assisted path

On Business or Enterprise, AI can clarify an idea into a PRD draft, break requirements into stories, and propose engineering issues. Pro and above can check and split stories. Inspect cost estimates when offered, request a draft, review its content and coverage, and explicitly save or create the selected work. PRD clarification can consume credits before its larger-request estimate is shown.

AI drafting does not replace PRD approval. An INVEST result or coverage indicator is a prompt to inspect the work, not proof that a requirement is correct or shipped.

Keep cost and scope deliberate

Start with a small set of requirements or stories. Cost estimates are previews, not a fixed-price guarantee; retries, follow-up drafts and duplicate checks can add usage. Workspace budgets and feature switches still apply. See AI plans and cost controls.

Using an agent or the API

The Velocity CLI and GraphQL API expose the shipped PRD and story operations. Use the current MCP tools reference to check which actions your MCP client supports today. Dedicated PRD/story planning MCP tools and a one-click AI pipeline are separate work; this guide describes the individual actions available now.