🧵
fray

A local dashboard for running many coding agents at once.
by @colinhacks

What is fray?

Run one agent and you babysit a terminal. Run ten and you lose track of which ones are working, which
are stuck, and which are waiting on you.

fray is a web UI that runs on your machine, scoped to one repo. You describe a task, fray dispatches a
real Claude Code or Codex
session to drive it, and every session shows up as a thread you can watch, steer, answer, or drop into
a live terminal. Threads that need you sort to the top; the rest get out of the way.

There is no cloud, no account, and no agent of its own — fray is a window onto the CLIs you already
have installed.

Quickstart

Requires Node 26+, tmux, and the Claude Code or
Codex CLI signed in.

cd /path/to/your/repo
npx frayui

That's the whole setup. It opens a browser tab, and one server runs per repo.

How it works

• One thread, one effort. Each thread owns a single top-level agent session in a detached tmux
  session, so closing the tab never kills the work. tmux -L fray attach is always there if you want
  the raw terminal.
• Agents signal, they don't just stop. A worker ends its turn with a fenced ```done,
  ```awaiting, or ```question block. fray renders those as cards — a question becomes
  answerable choices, an awaiting pr-watch: parks the thread and wakes it when the PR moves.
• Nothing is hidden in the app. All the orchestration judgment lives in text you can edit: the
  dispatch preamble in Settings, and a FRAY.md at your repo root whose contents are injected
  into every worker and override fray's defaults.
• State stays out of your repo. Unread counts, settings, and the session registry live in
  ~/.fray/projects/<id>/ui.db.

Docs

• ARCHITECTURE.md — the invariants, layout, and design decisions. Read it before
  changing anything.
• cc-worker/ — the Claude Code plugin every dispatched agent loads.
• board/ — the zero-dep .fray/ board parser the server shells out to.

License

MIT