First session

Open a repository

From a terminal, in or above the repository:

npx refyard open /path/to/repository

refyard open starts a service on loopback, prints a pairing URL, and opens it in your browser. The URL carries a single-use ticket: the browser spends it once and keeps an in-memory session token for that service process.

Nothing is copied or indexed ahead of time. Refyard reads the repository with the machine’s own git, in place.

`refyard serve --repo a --repo b` approves more than one root; each becomes a tab in the workbench and a row in the Repositories panel.

Pair another browser without restarting

A pairing ticket is single-use by design, and the service only mints new ones on trusted local channels — never over HTTP. To hand a fresh URL to another browser, ask the running service from a terminal on the same machine:

refyard pair              # prints a fresh pairing URL for the service on this machine
refyard pair --port 9595  # when more than one service is running
refyard pair --json       # machine-readable

refyard pair talks to the service over a same-user-only control socket, so it needs no terminal interaction and no restart. The p keystroke on the serving terminal does the same thing for whoever is sitting in front of it.

What you are looking at

Area	What it does
Repository tabs	One tab per approved repository, restored on reload.
History	The commit graph, refs, search and filters — see History.
Working copy	Staged, unstaged and untracked paths, with per-file diffs — see Working copy.
Panel rail	Branches, Remotes, Stashes, Tags, Worktrees, Submodules — see Panels.
Right panel	The selected commit’s details and diff, or the working copy’s commit box.
Status badge	Live-update state: live updates, connecting…, or no live updates.

Live updates arrive over an authenticated event stream, so the graph, the status and the panels follow the repository while it changes — including changes made in a terminal or an IDE next to you.

If something looks wrong

The workbench never guesses. When a read has no answer it says which read failed; when a write’s outcome is unknown it says so and refuses to continue with dependent writes. See Troubleshooting.