Provance Engine
From idea to a signed, traceable delivery package.
Engine takes a product idea through intake, research sprints, workshops and architecture, and compiles the package your engineers build from. You make the decisions at every gate. Every document records what it was made from.
How it works
A gated method, not a chat window.
Each phase ends at a gate. Provance presents its conclusions; you approve, reject or redirect them before the next phase starts.
- 01
Idea
Describe the product you want to exist, in your own words. - 02
Intake
Provance asks what it needs to know and records your answers. - 03
Research sprints
Structured research that challenges the idea and surfaces the decisions that matter. - 04
Workshops
Product and architecture decisions, made by you, at each gate. - 05
Architecture
Requirements, technology choices and sprint-level implementation specs. - 06
Delivery package
One compiled package with a signed manifest and its full record.Signed
What you get
The documents your team builds from.
- Product requirements document
- Architecture requirements, with the technology choices and the reasons for them
- Sprint-level implementation specifications
- One compiled, downloadable delivery package
- A signed manifest and the record of how each document was produced
Traceable from the first prompt
Every Engine artifact is recorded in a hash chain for its project, from the concept prompt to the signed package. Nothing is edited in place: a change makes a new version and the old one stays.
How the evidence worksYour privacy settings apply
Engine documents follow your organisation's choice for personal details, the same as every other Provance product.
Models
Use an Engine-grade model, yours or ours.
On the Free plan you bring your own key. Pro and Team include Provance-hosted AI, and you can still use your own key on any project with no Provance limit. One chosen model runs chat, intake and deliverables alike.
| Provider | Accepted with your own key |
|---|---|
| Anthropic | Claude Opus 5.5, Sonnet 5.5, Fable 5.1 |
| OpenAI | GPT-6 Astra, GPT-6.1 Sol, GPT-6 Sol |
| Gemini 4 Argon | |
| Z.ai | GLM-5.2 |
Small and fast models, older generations and some providers are not accepted for Engine work. The list is kept in the app and can change as new models are evaluated.
Plans
Start free. Pay for capacity when you need it.
You pay for projects at once and users, not per document. API and SDK access are included on every plan.
Free
Try the full method on one project with your own model key.
Free for 90 days
- 1 project at a time (finish it to start the next)
- 1 user
- Your own key for an Engine-grade model
- API and SDK access
Recommended
Pro
For a small team running several projects.
$249 /month
Plus GST
- Up to 5 projects at once
- 3 users
- Provance-hosted AI for about 7 project runs a month (about 6 on annual billing)
- Your own key for any project, with no Provance limit
- API and SDK access
Team
For delivery teams working across many projects.
$699 /month
Plus GST
- Up to 20 projects at once
- 10 users
- Provance-hosted AI for about 26 project runs a month (about 24 on annual billing)
- Your own key for any project, with no Provance limit
- Team management and priority processing
Enterprise
Contracted users, projects and capacity.
From US$40,000 a year
- Users and projects set by contract
- Up to 50 generations at once
- Support terms agreed with you
Hosted capacity is stated as project runs at the current hosted model. Organisations with a paid Synthetic plan get 10% off Engine. Full pricing.
FAQ
Engine questions.
Which models can I use with Engine?
On the Free plan you bring your own key for an Engine-grade model: Anthropic Claude Opus 5.5, Sonnet 5.5 or Fable 5.1; OpenAI GPT-6 Astra, GPT-6.1 Sol or GPT-6 Sol; Google Gemini 4 Argon; or Z.ai GLM-5.2. Small and fast models, older generations and some providers are not accepted for Engine work. Pro and Team include Provance-hosted AI, and you can still use your own key on any project.
What does an Engine delivery package contain?
The documents your engineering team builds from, including the product requirements, architecture requirements, technology choices and sprint-level implementation specifications, compiled into one package with a signed manifest and the record of how each document was produced.
Take your next idea through the method.
Free for 90 days on one project with your own Engine-grade model key. Upgrade when you want hosted AI or more projects.