Sometimes the agent needs material that isn't in the repo: a sample CSV, a failing request payload, a design spec, a screenshot of the bug. The Extra Context tab is where that lives — readable by the agent, separate from your code.

What it's for

Each env has an Extra Contexttab backed by a folder that sits alongside your repos but isn't part of them. Use it for reference material the agent should see but that doesn't belong in version control:

Data files — CSVs, JSON fixtures, sample exports.
Reproduction material — a captured request/response, a payload that triggers a bug.
Specs and notes — a feature brief, acceptance criteria, a design doc.
Screenshots and images.

Adding files

The tab is a file tree. You can upload files and whole folders (the folder structure is preserved), create files and folders inline, edit text files in the built-in editor, and rename, move, or delete entries. It's built for real working sets — thousands of files and large uploads are fine.

How the agent uses it

The Extra Context folder is handed to the agent as an additional working directory, so it reads the files natively — the same way it reads your code. Point it at a file in chat ("the agent can use extracontext/customers.csv") and it just works. The agent can also write here, which is where deliverables come from.

Where the agent puts its output

When you ask the agent for a report, a spec, or an export, it saves the result under extracontext/ai/. It can even render Markdown to a PDF there, which then shows up in the tab ready to download — a clean way to get a polished artifact out of a session.

Extra context vs. assets

Don't confuse Extra Context with an env's assets. Assets are files that are part of how the env runs — a docker-compose.yml, a schema.sql mounted into a database, an nginx.conf — provided when the env is created and referenced from the compose. Extra Context is reference and deliverable material that has nothing to do with the runtime. Both live on disk in the env; they serve different jobs.