Products · Wherever you work

The answer from wherever you are.

All of it works where your team already works: WhatsApp, Teams, email, the AI tools you use today, or our own product, Meridian. The agents work there too, so nobody has to learn a new system.

01 · The doors

Every door, the same brain.

WhatsAppAsk the brain a question, or send an agent a job, from the phone in your pocket. The answer comes back in the thread.

TeamsThe same, inside the workspace your office already runs on.

EmailSend an agent a request; the document arrives in your inbox when the work is done.

Claude and ChatGPTThe connector: the brain reached from inside the AI apps your people already use, under the same permissions.

Any other appAnything that speaks the open standard for connectors. Ask us and we add it.

MeridianThe full picture in one place, and where you define and manage your agents.

02 · The connector

Inside Claude and ChatGPT.

Your people already use Claude or ChatGPT. The connector lets them put any question to the whole brain from there, answered by the same tools and safeguards as Ask. It is built on the Model Context Protocol, the open standard both apps use for connectors.

How it connectsOne web address, pasted into Claude or ChatGPT. People sign in with their normal Kadura account and their usual authenticator code. No new password, no API key, no shared token. One consent screen, the first time only.

What it reachesEverything in the brain the person is allowed to see, with the definitions that make it make sense, so the app reads your figures the way your finance team does. A complete listing as an Excel file when you need one.

It reads, never writesNothing reached through the connector can change or delete your data. The one exception is Excel export, which writes a new file into your own storage.

The same permissions, exactlyThe connector is not a second door with its own rules. What a person can see in Meridian is exactly what they can see through Claude or ChatGPT, and nothing more. Every request runs as that person, under their own role, their own permissions and their own document clearance. If a set of data is closed to them in Meridian, it stays closed here, and a document above their clearance is never returned.

Under controlWhich apps may connect, which roles may use it, and an off switch that takes effect within a minute, all in your control plane, and in your own administrators’ hands once your deployment grants it. A person's access can be withdrawn on their next request. Every request is logged: who, which tool, which app, when. Never the question itself.

Agents, tooAny agent can be connected on its own, carrying its instructions, its private files and its skills into Claude from the first message. One sign-in covers all of them, and disconnecting one leaves the rest untouched.

What travels with every figure

The filters applied, how many rows it rests on, whether one row dominates, whether the list is partial, and a check verdict: corroborated, self-consistent, contested, single-source, or settled by a person. The honesty rules are handed to the app at the moment it uses a tool.

Where it stops

Answers go to the AI app or the channel your people use, under that provider’s terms, and only approved apps can connect. Kadura supplies the data, the rules and the checks. Claude or ChatGPT writes the final sentence.

Live in production with both Claude and ChatGPT. If your people work in any other AI application that supports connectors, ask us and we add it, under the same permissions and the same controls.

Included with every brain, and switched on when you ask. Switching it on is configuration, not a project.

ToolingBuilt for your operation alone

Tell us where your team already works.

Tell us about your business The control plane