DryDock

Case study · Tech & fun

Dispatch coding agents into isolated worktrees, stream runs to a phone PWA, and inspect test results, usage budgets, project documents and backlog. Runs locally with CLI subscription logins.

Stack: Next.js 15 + better-sqlite3 + Node subprocess spawn; SSE streaming to a mobile-first PWA over a Cloudflare Tunnel.
Status: In Progress
Dates: Reviewed October 2, 2026; data dates shown in the app

Background

How can a phone follow coding-agent work running on a local Mac?

DryDock dispatches CLI coding sessions into separate worktrees and streams run output to a mobile PWA. Subscription logins remain with the CLIs; tasks carry model, autonomy and queue settings.

The local dashboard now leads with command-center status, a suggested next action and weekly activity. It also exposes project documents, cross-project backlog and provider usage. Public screenshots use demo projects; no public deployment is available.

Screenshots

Demo data captured October 2, 2026. This app runs locally.

The dashboard brings together command-center status, a suggested next action, weekly activity and projects (mobile) — The dashboard brings together command-center status, a suggested next action, weekly activity and projects. This capture uses isolated demo responses.

The dashboard brings together command-center status, a suggested next action, weekly activity and projects (desktop) — The dashboard brings together command-center status, a suggested next action, weekly activity and projects. This capture uses isolated demo responses.

One-tap session kickoff: a free-form prompt against a chosen project, Claude or Gemini, and per-run model and autonomy overrides (mobile) — One-tap session kickoff: a free-form prompt against a chosen project, Claude or Gemini, and per-run model and autonomy overrides — dispatched on the CLI's own subscription login, no API keys anywhere.

One-tap session kickoff: a free-form prompt against a chosen project, Claude or Gemini, and per-run model and autonomy overrides (desktop) — One-tap session kickoff: a free-form prompt against a chosen project, Claude or Gemini, and per-run model and autonomy overrides — dispatched on the CLI's own subscription login, no API keys anywhere.

Features

Dispatch: Session kickoff selects a project, prompt, provider and per-run settings.
Isolation: Separate worktrees contain agent edits, with a quality gate after a run exits.
Run output, progress and usage appear in the PWA.
Operations: Queue, cancellation, resumable sessions and usage windows support ongoing work.
Local access: Loopback binding and configured remote authentication protect the local service. Diff-to-merge remains separate future work.

Reviewed October 2, 2026.

← Back to Homepage