SprintBoard turns one paragraph — a product launch, a half-marathon, a move to Lisbon — into a phased sprint of real goals. LLMs plan. Agents execute. You stay in review.
Most boards start with tickets someone already knows how to break down. SprintBoard starts with the outcome and figures out what the tickets even are.
The whole product fits in three verbs. Everything else is chrome.
One paragraph. Constraints, deadline, context — whatever you have. No taxonomy, no custom fields, no forty-field onboarding.
Phase 1 = quick wins. Phase 2 = structural work. Phase 3 = launch. Every goal gets a type, priority, estimate, dependencies, and acceptance criteria.
Assign a goal to a Claude-Code agent — it picks up the branch, makes the change, opens review. Your job is approve / reject / re-plan.
Same board. Same phases. Same agents. The only thing that changes is the paragraph you typed.
Features earn their place. Every one of these was missing from the boards we already use.
Quick wins → structural → launch. The plan respects dependencies and always surfaces the next best goal.
Assigning a goal to an agent is a primary verb, not a plugin. Agent chips live next to human avatars.
Done stays, open gets re-phased against what's left.
Reviewers actually know what “done” means.
⌘K · J/K · A · R · Esc. Everything else is slow.
Claude, GPT, local. The model is a preference.
SprintBoard is in early access for solo builders, small teams, and anyone tired of empty boards.