Write the scope. Your AI team ships it. You hold the gates.
Write the scope in one sentence. The PM breaks it down, the analyst writes stories and acceptance criteria, the architect drafts the technical plan; the developer codes on its own branch, the architect reviews, QA tests, and the pull request merges when CI is green. You only decide at the gates.
Your company works in its own space at yourcompany.gokteams.com: your projects, agents, documents and costs are visible to your team only. Teammates join by invitation; nobody gets in uninvited.
Every item flows through a fixed pipeline; each stage produces an editable document (breakdown, risk report, business analysis, technical plan, change set, review, test report).
Connect your GitHub projects; the Curator reverse-analyses each repo into a code map, conventions and a technical debt list. Open pull requests and issues are imported.
Risk thresholds, item size and review-loop policy are set per company. High-risk scope, large items and release decisions go to a human; the rest is automatic.
Instruct any role from the team chat, edit a document, pause an agent and write the code yourself in the in-browser IDE, then let it continue.
Files, branches and pull requests, GitHub Actions pipelines, deployments, releases, one-click rollback, uptime monitoring and error signals on one screen.
Decisions, conventions and pitfalls from every finished item go into the knowledge base, so each item is cheaper and more accurate than the last.
Tokens and dollars per item, usage by role and model, human-hour equivalent. Bring your own API keys (BYOK) or subscriptions.
Every role is a separate agent; its model, persona and capacity are managed from the HR screen and compared with A/B runs.
Your code and data stay yours; the factory acts only with the access you grant.
Every user signs in with a password or e-mail link plus an authenticator app code (TOTP); the step cannot be skipped.
Every read and write is locked to your company, and the rule is checked automatically on every code change; an unscoped query cannot ship.
The GokTeams team cannot see your workspace. For an issue you grant time-limited, read-only access; you are notified on every entry and can revoke it any time.
Agents work on their own branches and open pull requests; nothing reaches the main branch outside CI and your policy. A breaking change is rolled back in one click.
Viewer, operator, admin and owner roles; gate decisions, settings and membership stay with authorised people.
Model provider keys you bring are stored encrypted and used only for your own work.
You choose the number of users: every person with access is a licence, agents are not counted. Add licences as your team grows, remove them as it shrinks. Want everything unlimited? Choose Premium. AI usage is paid through your own keys or subscriptions, with a spending cap you set.
It runs GokTwins' own products first; a pilot programme is open for other companies. Prices will be published here once final.
Everyone who signs in: viewer, operator, admin or owner. Agents and automated processes do not count. On the Standard plan you can add or remove licences at any time.
In your own GitHub repository. Agents open the repo in a temporary working copy and send changes as branches and pull requests; merging follows your policy and your decision.
Chosen per role: Claude, Gemini and OpenAI models are supported, and new ones are added to the model catalogue without code changes. If a provider hits its limit, thinking roles can continue on a backup model.
The architect reviews every change, QA runs the tests and CI results are visible. If a loop gets stuck, the item is not silently left hanging; it comes to you with an explanation. A bug that reached production is rolled back in one click.