Context is everything.

Yes, it is designed to fit into a validated process. The AI only drafts: every requirement and every decision is approved by a named person before it counts, and only approved content becomes the record. Every change is stored as a commit, so you can always see who changed what, and when. That gives your quality team a clear intended use to validate against: AI-assisted drafting and conflict checking, with people making every decision.

Approved requirements are stored as plain files in a git repository that you control, and that repository is the source of record. Because they are plain files, you can read, back up or move them without Centropic, and because every change is a commit, you get a complete history. Centropic creates and updates the linked tickets in Jira and Linear, so your teams keep working where they already work, and teammates can read and search requirements in the web app.

Centropic reads each file to pick out the requirements, rules and decisions it contains, and then deletes the original. What stays is what Centropic needs to check future requirements against: the extracted requirements, the decisions and a search index. All of it stays in your workspace, hosted in the EU.

No, and no tool can do that on its own. Compliance comes from your quality process, your people and your evidence. What Centropic does is make that evidence easier to keep: who approved each requirement and when, the decision that resolved each conflict, and the links from requirements to risks, test cases, work items and releases. When an auditor asks why something was built a certain way, the answer is already written down.

Yes. Some teams write user stories, some write "shall" statements, and many have their own template with the fields their process requires. You decide how requirements are worded and structured, and the assistant drafts in that format, so new requirements look like the ones your team already writes.

Yes, on request. Some organisations, especially in regulated industries, need to review the software their requirements live in, for example as part of a supplier assessment or tool validation. If yours does, ask us and we'll make the source code available for review.

Yes. Centropic follows the permissions you already have in Jira, Linear and Confluence, so people only see the requirements and sources they could already open in those tools. Restricted content is still checked: if a new requirement conflicts with something the writer isn't allowed to see, Centropic tells them a conflict exists, without showing the restricted content, and asks someone with access to resolve it. The conflict is caught and nothing leaks. AI agents connecting through the MCP server get no more access than the person using them. Roles per project, single sign-on and an audit log are next on the roadmap.

The first release connects to Jira, Linear and Confluence. Centropic reads what is already there, such as tickets, pages and existing requirements, to write new requirements and check them for conflicts, and it keeps linked tickets in Jira and Linear up to date. You can also upload notes and documents or paste text. AI coding agents, such as Claude Code, connect through Centropic's MCP server. GitHub and more sources are planned on the roadmap.