Projects and workspaces

A project is a folder on your machine. It gives Raticode a place to read files, run commands, and register workflows.

Open a folder

Choose Open Project (File > Open Project..., or Open project... in the project dropdown) and use the native folder picker. Recent projects make it easy to return to folders you have already opened. The active project appears in the workspace controls; check it before running a workflow or terminal command.

Raticode can keep editor files and workflow tabs open while you move between projects. An open tab is a reference to a file, not a request for Rem to inspect it.

Where workflows live

Registered workflows live beneath their project:

my-project/
  .raticode/
    <workflow-id>/
      workflow.rattish
      workflow.metadata.json
  src/
  reports/

workflow.rattish is the executable source. Metadata records graph positions and editor state. Compiled artifacts and run records are managed internally. Edit the source through the studio or CLI; do not hand-edit compiled IR to change behavior.

Use Raticode rattish list --format json to inspect registered workflow paths. Copy the actual source path when scripting instead of constructing one from a title.

Rem has its own scope

A project-scoped thread keeps its project when you change the active editor. Check the scope shown in Rem before requesting a modification. A global thread can research across available resources and select an open project for project work.

Scope selection does not clone a conversation or change its identity. Provider and model choices also remain part of the same thread. See Chat and threads.

Filesystem grants

The desktop uses selected folders and files as grants for its file operations. A native folder picker is the normal way to grant access. Opening a project does not install its dependencies or authenticate its providers.

Provider sandboxes, workflow path rules, and desktop file grants have different responsibilities. Avoid interpreting the resource picker as a restriction on every process the OS user can start.

For isolated parallel Git work, use worktrees in Source control or Swarms.