The ones a cautious buyer actually asks — about your code, your data, and what the AI is and is not allowed to do. If yours is not here, ask us.
No, and there is no code path that could. Corporate AI 365 reasons over source code and scripted database objects only. If live data is genuinely needed to diagnose something, the AI writes a read-only query for your own developer to run.
No. It opens branches and pull requests, and production is only ever reached by a merge that your own approver authorises. Your branch protection, reviews and CI still apply.
No. The platform orchestrates your delivery process; it never hosts or runs your application. We merge a pull request and your own CI deploys, exactly as it does today.
Every record is stamped with its tenant and filtered at the database layer on every query, rather than by each feature remembering to add a condition. It is enforced centrally so a new feature is isolated by construction.
Yes. Analysis runs at temperature zero and the result is cached against a hash of the issue and the code snapshot it was analysed on. Asking again returns the stored analysis rather than calling the model again.
Not for the trial, where AI is included. On paid plans you can use the platform quota or bring your own OpenAI or Anthropic key, whichever suits your procurement. A key of your own is billed by the provider directly and is not counted against any platform allowance.
Not by us. Your code is sent to the AI provider you are configured against — Anthropic or OpenAI — solely to answer your own request. Their enterprise API terms govern what they may do with it, and if that matters to your procurement, bringing your own key puts that contract directly between you and them.
GitHub, GitLab, Bitbucket and Azure DevOps. Each is a connector behind one interface, so the fix pipeline, promotion and ingestion behave the same whichever you use.
Yes, as one pull request bundling every fix that shipped together. Every file is checked first, and if any has been edited since the release the revert refuses and names it rather than overwriting somebody's work.
Not as a like-for-like swap. Task management here is deliberately a subset — boards, sprints, tasks and sub-items attached to the work the AI is already reasoning about. Teams running heavy Jira workflows should expect to keep them; teams drowning in tool-switching usually find this is enough.
Your company domain is how we group your team, let colleagues join, and tell staff apart from external guests. That does not work on a personal mailbox, so trials start from a work address.
Your account becomes read-only. Nothing is deleted and you keep access to everything you created. Pick a plan whenever you are ready and it unlocks immediately.
Two things, metered separately: embedding, which is dominated by importing a repository, and analysis, which is the reasoning behind an answer. Importing is by far the larger of the two. You can see the current period's usage in your billing screen before you reach any limit.