AI context, the notes your team writes for agents

An agent that knows your conventions writes code your reviewers accept. The AI context is where you write them down: free Markdown for AI agents, one for the project and one for each microfrontend.

Context	Where to edit it	What belongs there
Project AI context	Project settings → AI context tab	What applies to every microfrontend: the domain of the product, shared conventions, rules
Microfrontend AI context	The AI context tab of the microfrontend	Its responsibility, its quirks, what not to touch

Each is Markdown, at most 20,000 characters.

## Domain
Customer portal for the B2B shop.

## Conventions
- Strict TypeScript, no `any`
- Shared UI components come from the `design-system` remote: never copy them
- Every user-facing string goes through i18n, in Italian and English

Who reads it

An agent always gets the context of what it works on, plus the project one:

• Edit with AI and AI jobs put the project context and the context of each microfrontend in the agent's prompt, together with the stack detected in the repository.
• Maestro reads the same notes through the same tools.
• MCP clients are asked to call context_get before working on a project. It returns the project context, and the microfrontend one when given a microfrontendId; microfrontend_get returns the same notes in aiContext, so a client reading a microfrontend gets them without a second call. project_update and microfrontend_update change them.

Keep the project context short and general, and let each microfrontend say what is specific to it: agents read both.