Stackship Studio

Formerly Circuit. The film and screenshots still show the old name.

Early access · Pilot program

AI agents build. You decide at every step.

Stackship Studio builds business systems on your Stackship platform. You describe the need in your own words, the agent builds the system in the conversation, and a person classifies it and puts it into production.

  • You choose the models and providers
  • A person puts the system into production
  • The code and documentation live in your repo
Animation: a need becomes a lunch-ordering app that is built in the conversation, classified and confirmed by a person in Finish, and runs in Operate

Studio · from need to production

How it fits together

Studio runs on Stackship

Studio is where systems are built and maintained. Stackship runs them, and the resources, sign-in, price list and AI profiles come from the platform.

Studio: where systems are built

The conversation with the agent, the prototype, the forms in Finish, the decision log and the operations view in Operate.

Stackship: where systems run

Apps, functions, databases, queues and secrets in an isolated environment, with the platform's sign-in and billing. Sentinel reviews the cluster and shows its findings in Operate.

Studio creates the resources in Stackship as the signed-in person. The isolated environment is chosen when the repo is created or when the system goes into production, and cannot be changed afterwards.

Three steps from need to production

Build, Finish and Operate. The next step opens when the previous one is done.

  1. 1

    Build

    You describe the need in your own words and the agent builds a prototype in the conversation. It asks what it needs to know and offers answers to choose from. Every round of work produces a new version you can try at once and roll back to.

    • Preview at desktop, tablet and mobile width
    • Text, images and styling can be edited directly in the preview
    • Once the prototype works you pick a runtime and a database, and the agent builds the production version
    Build · lunch ordering, prototype in the conversation
  2. 2

    Finish

    The agent fills in the forms from the conversation and asks for what is missing. A field nobody has answered stays unanswered, and the system cannot go into production until everything is filled in.

    • Classification of confidentiality, integrity and availability on levels 1–4 and of personal data, confirmed by a person
    • System documentation, a runbook with backup, RPO and RTO, architecture decisions and an agent contract
    • The resources, with a monthly cost from the platform's price list
    • Named owners for operations, development and product
    Finish · the agent's proposed classification
  3. 3

    Operate

    Going to production creates the resources in Stackship, and the platform builds the system from your repo. Once it runs, Operate shows its address along with operational data and the full record.

    • Load, logs, alarms, backups and deployments per service
    • Sentinel reviews the cluster's configuration and logs every 30 minutes, and dependencies are scanned for vulnerabilities daily
    • A change gets its own branch and conversation, and goes in when a person merges it
    Operate · to-do list in production

What the agent builds with

An administrator sets up models, sources and design systems once. Every initiative then uses them.

Models per task

Building, review, summaries and Swedish wording can each go to a different model. Every model is approved for a confidentiality level, and Studio can use the AI profiles in your Stackship installation.

Knowledge sources

Upload documents or read in a website. Each source has a confidentiality level, and each initiative picks the sources the agent may read.

Design systems

Colours, typefaces, components and logos for the agent to build with. A design system can be generated from a description or from your brand material, and chosen per initiative.

Review

A separate model reviews code and security and leaves findings in Build. For each finding you choose whether the agent fixes it, you fix it yourself, or it is not a problem.

Insights

Ask questions of connected systems and get the answer as a report you can save as PDF or share in the organisation. An answer that points to a need can become a new initiative.

Decision log

Every decision is recorded with who made it and when. The log is a hash chain verified each time it is read, and the forms show whether a person or the agent filled them in.

The principles behind Studio

What governs how the agent works and what it may do.

People decide

The agent writes the documentation and proposes the classification and resources. Confirming the classification, putting the system into production and merging a change take a signed-in person.

Gaps stay visible

The agent only fills in what the conversation has answered. Anything missing shows as unanswered, and going to production stays blocked until the list is complete.

Business language

The agent asks what the system should do, for whom and at what cost. Frameworks and technology come up when someone raises them.

Your code

The initiative's repo is created in your GitHub or GitLab and Studio delivers to it after every round of work. The systems run on Kubernetes and can move to another cluster.

The record lives in your repo

Every form and decision is committed as a markdown file in the initiative's repo, where whoever reviews or takes over the system will find it.

  • The needdocs/need.md
  • Classificationdocs/classification.md
  • System documentationdocs/system.md
  • Runbookdocs/runbook.md
  • Architecture decisionsdocs/adr/
  • Agent contractdocs/agent-contract.md
  • Resourcesdocs/resource-manifest.md
  • System ownersdocs/owners.md
  • Decision logdocs/decision-log.md

Questions

What buyers tend to ask first.

The ones your administrator adds: from a provider with an OpenAI-compatible API, from the AI profiles in your Stackship installation, or running locally on your own hardware. Every model is approved for a confidentiality level, and the agent only reads knowledge sources up to that level.

Pilot program

Join the Studio early-access pilot

We're working with a small group of public-sector and regulated organisations to shape Studio before general availability.

Who it's for

  • Public-sector bodies: agencies, regions, municipalities
  • Regulated industries with requirements on where data and models live
  • Organisations that want to build their own systems without waiting for a developer
  • Organisations already running Stackship, or planning to

What you get

  • Studio on your own Stackship platform
  • Direct contact with the team building Studio
  • Early-access pricing

Studio is in active development. Pilot scope and availability are agreed per organisation.

SEE IT LIVE

Request early access

Tell us what you’d like to explore and leave your email, and we’ll be in touch about early access.

What are you interested in?

Studio runs on Stackship, so it’s included automatically.

No commitment, just a conversation and a walkthrough.

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.