Repository features

Issues, projects, releases and «Learn» are switched on and off with checkboxes — the repository adapts to what it is for. AI has its own section.

Who can use this Features can be changed by a repository admin.

Updated Copy as Markdown
In this article

Where#

Settings → Features — a separate item in the settings column, like Features on GitHub.

What can be turned off#

Feature What gets hidden
Issues the «Issues» tab and issue pages; an issue cannot be created either on the page or over the API
Projects the «Projects» tab; a project cannot be linked to the repository
Releases the «Releases» tab and link; releases cannot be published or deleted
Learn the «Learn» tab, drills and challenges from notes through the agent, this repository's challenges in drills and review
AI (the AI section) the repository chat and generating questions and tasks from notes

Pull requests cannot be turned off: they are part of working with git, and GitHub does not let you turn them off either.

What happens to the data#

Nothing is deleted. A disabled feature is only hidden: issues, releases, challenges and the review schedule stay in the repository. Turn it back on and everything is in place, and reviews continue from where they stopped.

Over the API a disabled feature answers 410 with the REPOSITORY_FEATURE_DISABLED code.

Defaults#

A new repository has everything on. Private repositories whose owners had not allowed AI to read private content before 11.10.2026 got AI turned off. Turn off what this repository does not need: for example, «Learn» for a code repository, «Releases» for study notes.

AI#

The AI section of the repository settings: whether AI is enabled here (when off, Space AI does not read this repository and does not generate tasks) and how much AI spent in it this month. Which key pays is chosen in the chat — see Connecting models.

Agents#

The same section has Agents: what Space AI and external MCP agents may do in this repository (read only; read and manage issues; everything — also changes through pull requests and notes) and the context repository — the project memory the agent reads together with this repository. You can pick a repository you also manage. In detail — MCP server.

Was this helpful?