What changed: Core is now available inside the AI surfaces and work tools your team already uses.
Gradien is launching Core, a layer that sits underneath the AI tools your team already opens every day. Core connects to surfaces like ChatGPT, Claude, Cursor, Codex, and Claude Code, along with the work tools where decisions actually get made and recorded: Gmail, Google Drive, GitHub, Notion, Slack, Google Calendar, Microsoft Teams, and Linear. Once connected, Core captures the context behind completed work, with your approval: the information, decisions, preferences, corrections, actions taken, and outcomes that normally live in someone's memory of a project rather than in any single document. That becomes proprietary knowledge Core makes available the next time you or a teammate opens any of those AI surfaces.
Core is not a separate chatbot you have to visit, and it does not replace the AI tools your team already relies on. It is a continuity layer that follows the work across whichever surface you happen to be using that day, so the assistant in front of you already knows what has been decided, why, and what happened as a result.
Why it matters: the context behind your team's work has been getting lost between tools.
Most teams now move between several AI surfaces in a single week, sometimes in a single day. A decision made in a Claude conversation on Monday does not show up when someone opens ChatGPT on Wednesday to continue the same project. A correction one teammate gave an assistant in Cursor does not travel to the assistant a colleague opens later in Claude Code. Each surface starts from the same blank page, and someone has to re-explain the background before real work can continue.
The work already happened somewhere. Core's job is to make sure the next conversation does not have to start over.
That re-explaining has a cost that is easy to overlook because it is spread across many small moments: a paragraph of background pasted into a prompt, a Slack thread searched for the reasoning behind a decision, a teammate asked to repeat something they already explained last week. Core is built to reduce how often that repetition is necessary, by keeping the reasoning attached to the work instead of leaving it trapped inside a single conversation.
How it works: Core connects, captures with approval, and retrieves knowledge wherever you're working.
Core connects to the tools where decisions and work already happen.
You connect Core to the AI surfaces and work tools your team uses day to day. Core reads across those connections to understand what a piece of work involved, not only where the final file ended up.
- ChatGPT, Claude, Cursor, Codex, and Claude Code
- Gmail and Google Calendar
- Google Drive and Notion
- GitHub and Linear
- Slack and Microsoft Teams
Core captures the workpath behind finished work, not only the final output.
Consider a team like Maya, Jordan, and Alex working on a product launch for a client such as Meridian Labs. Over a few weeks they make dozens of decisions: which messaging to use, why an earlier draft was rejected, what the client corrected on a review call, what launch date finally got approved after two changes. A finished document only shows the last version. Core is built to capture the path that led there, the drafts, the corrections, the reasoning, and the outcome, so that the next person, or the same person three months later, can see not just what was decided but why.
Company knowledge only changes after a person approves it.
Core does not retrain or fine-tune any of the underlying AI models, and it does not change what counts as company knowledge on its own. When Core captures something worth keeping, a person reviews and approves it before it becomes part of the knowledge Core makes available. That approval step is intentional. It keeps a team's proprietary knowledge accurate and accountable to the people who did the work, rather than treating every capture as automatically correct. Once something is approved, it becomes available inside whichever AI surface you use next, so the same decision does not have to be typed into a prompt twice.
Related resources: read more about how Core fits into your AI workflow.
- How Core captures a workpath, not just a final document
- Why proprietary knowledge should not depend on which AI tool you happen to open
- Setting up Core across ChatGPT, Claude, and the tools your team already uses