Org-Level Intelligence

After mapping repos and binding documents, connect a model provider once for the org so people can ask questions in plain language. Publish boards when people only need read-only access without an account.

This page explains what needs a provider and what works without one.

Important

New here? Build a landscape first: the Plugin or Chat path builds one from an interview, the Hands-On path has you draw it.

What runs without a model, and what needs one

The plugin loop needs no provider. The code plugin runs inside your AI host. /analyze builds the graph locally, /sync pushes it, and /adopt extracts an aspect from the same session.

The connect plugin also needs no provider. Its /sync checks an authored board against a document repo, proposes evidence links, and reports drift. Neither plugin calls a model on our servers.

Two things need a provider: chat and an analysis started inside the portal. Both run on our servers, and our servers do not include a built-in model.

Runs with no providerNeeds a connected provider
The code plugin's /analyze, /sync, /adopt, /intents, /skillsChat, on every board screen
The connect plugin's /sync, and /insights from eitherAn analysis started from chat
Publishing, and everything a stakeholder readsIntents chat drafts for you

Connecting a provider

One connection covers the org. Open Settings → Connectors (/settings opens the same page) and authorize a model provider such as OpenRouter. Every workspace uses that connection; nothing repeats per board or person.

One org-wide connection enables chat for every workspace.
Tip

Deep dive: Integrations Reference — every connector and its authorization process.

What chat adds

Once a provider is connected, a Chat button appears on every board screen. One conversation follows you between boards and reads mapped nodes, edges, aspects, bound sources, and past insights.

Ask chat a question and watch it generate a focused context board from your landscape

Chat can create a context board, start an analysis, and draft intents. Drafted intents stay drafts until a person hands them off to developers. The scope chip shows how much of the landscape the answer used.

A new chat thread offers quick starts such as assessing the board or drawing a context board. Each quick start only pre-fills the message; chat chooses the tools from your request. Chat covers the full process, and Insights explains where insights appear.

Publishing to everyone else

Publishing needs no model. Publish a board and stakeholders can read it at a stable URL without an account. Choose the address, included insights, aspects, and skills, and visibility: public, unlisted, or protected. Re-publish to update; the URL stays the same.

Stakeholders read the board themselves, with no architect involved.

Publishing Boards explains the wizard, the visibility tiers, and what a stakeholder gets.

What's next