AI Generation
Turn a plain-language prompt into a complete project plan — milestones, tasks, estimates, stack, and metrics.
AI generation is the fastest way to start. You describe the work; sprintrr produces a structured, editable plan.
Start from an idea
If you're not sure what to type, the generate modal shows a row of starters — one-tap ideas spanning common project types (app MVP, marketing launch, event, research study, content, hiring, fundraising, process improvement, personal goal, and more). Pick one and it fills the prompt box with a solid starting description that you can edit before generating. The AI then writes a plan tailored to your edits — starters are a head start, not a fixed template.
Inputs
| Input | Description |
|---|---|
| Prompt | Plain-language description of what you want to build. Specificity improves the plan. |
| Project type | The kind of project — e.g. General Project, Software Development, SaaS Platform, Mobile Application, API / Backend Service, E-commerce Store, Data & Analytics, Marketing Campaign, Business Operations, Research & Analysis, Change Management. |
| Duration | Target timeframe, in weeks. |
| Complexity | beginner, intermediate, or advanced. Higher complexity asks for a deeper plan and takes longer to generate (see below). |
| Mode | Quality (default — Claude Sonnet 4.6, highest fidelity) or Speed (Gemini 3.5 Flash, ~4× faster output). Picked per generation; applies to the sprintrr-credits path only. |
| AI source | Which AI runs the generation — sprintrr's credits or one of your saved BYOK provider keys. Picked per generation. |
Complexity and expected duration
The complexity dial scales both the depth of the plan and how long generation takes. The modal shows an estimated time before you submit:
| Complexity | Expected duration |
|---|---|
beginner | ~10–30 seconds |
intermediate | ~30–60 seconds |
advanced | ~1–2 minutes |
There is no plan gate on advanced — every plan can use it. Pick the one that matches the depth of plan you want; you can always regenerate.
Choose your AI source per generation
When you have BYOK enabled with at least one saved provider key, the generate modal shows an AI source picker so you can choose, on each generation, whether the call runs on sprintrr credits or on your own provider key (Anthropic, OpenAI, or Gemini).
The chosen source — including the provider and model — is recorded in the activity feed on the generated project, so you always have an honest record of which AI produced which plan.
Choose Quality or Speed
A small Quality | Speed toggle sits next to the Cancel and Generate Project buttons in the modal. It picks which AI model runs your credit-path generation:
| Option | Model | Best for |
|---|---|---|
| Quality (default) | Claude Sonnet 4.6 | Production plans you'll actually ship — highest fidelity on nuanced prompts. |
| Speed | Gemini 3.5 Flash | Rapid iteration, brainstorming variations, or any time you just need a structural skeleton fast — roughly 4× faster output. |
Both options cost 1 credit. The choice is about latency and output style, not pricing — and both run on sprintrr's keys, so you don't need BYOK to use Speed. The model that actually ran is recorded in the activity feed alongside the source.
When BYOK is on, the Quality/Speed toggle stays visible but becomes inert (both buttons dimmed, disabled) — your BYOK model is already the source, so the mode picker doesn't apply.
What it produces
A complete project, generated and streamed back in real time:
- 4–6 milestones structuring the work into phases.
- 8–15 tasks (the count scales with type/complexity) with category, priority, and estimated hours.
- A suggested tech stack.
- Success metrics describing what a good outcome looks like.
- A total estimated-hours roll-up.
Generation streams progressively, so you see milestones and tasks appear as the model works rather than waiting for one final response.
Stopping a generation
A generation in progress can be cancelled. Clicking Stop during generation opens a confirmation dialog ("Stop generating?") to prevent accidental cancellation — confirm and the run is aborted; nothing is saved. Closing the modal without confirming keeps the generation running in the background.
Credit cost
Each generation consumes one AI credit from your balance — unless you have Bring Your Own Key enabled, in which case generation runs on your own provider key and does not consume sprintrr credits. Credit balances and plan limits are covered in Plans & Credits.
A focused prompt ("a Next.js SaaS for invoicing freelancers, Stripe billing, 8-week MVP") yields a sharper plan than a vague one ("an app"). You can always regenerate or hand-edit.
Templates
Projects can be exported as reusable templates and used to seed new projects with a known structure (milestones and tasks reset to Not Started, with relative date offsets recalculated from the new start date). See Data & Export for the template format.
After generation
The plan is a starting point — every field is editable. Adjust estimates, set due dates, regroup tasks under milestones, and start moving work on the board. See Tasks and Milestones.