One inbox, grouped by project
Every running agent in one column, with unread badges per project so nothing finishes unnoticed.
Shipwright · v0.6.3 is out →
Every coding agent.
One calm inbox.
Shipwright turns your terminal into a vertical inbox of projects and surfaces the local coding agents that need your attention. Agents waiting on input rise to the top; everything else stays quiet.
Signed & notarized · macOS 13+ · Apple Silicon + Intel
Windows 10/11 x64 installer · Alpha · Not code-signed yet, so SmartScreen warns on first run · Uninstall it from Settings → Apps
Claude CodeCodexAny local CLI agentNo accountLocal-only usage stats
Drives the local agents you already run
A quiet place for loud agents. Shipwright watches every session across every project and only speaks up when an agent needs a decision from you.
01Inbox
Parallel by default
Run a dozen agents across projects. The ones waiting on you rise to the top.
02Local
Local, not cloud
Hooks run on your machine. No relay, no sign-in, and no session ever leaves it.
03Resume
Picks up where you left off
Quit, restart, come back tomorrow. Every conversation reattaches with history intact.
1.0 Inbox
Parallel sessions
without the noise
Kick off as many agents as you want across every project. Shipwright pulls the ones that need a decision to the top with a status dot. When everyone is heads-down, it simply reads “Nothing needs attention.”
Token countsReTokenAsking
Markdown tablesImplicitDraftApprove
Plan putting fixesRangentRunning
Fairway centerlineRangentDone
Price feed retryReToken3d
Manifest schemaImplicitDraft1d
Sorted by who's waiting on you
Live Asking, Approve, Running and Done states. Quiet by default.
2.0 Local agents
Your machine.
Your agents. Your keys.
Shipwright drives Claude Code, Codex, and other local CLI agents through standard terminal hooks. Nothing is brokered through a server, and there's no account to create.
Bring your own CLI agents
Run Codex, Claude Code, or anything that speaks a terminal, with your own credentials.
$ printf '\e]777;notify;Asking\a'
Token countsneeds input↵
Standard hooks, OSC 777
No cloud relay, no sign-in, nothing leaves your machine. The terminal tells the inbox; that's the whole protocol.
3.0 Usage
Know what your
agents cost
The Usage page totals tokens, cache reads, sessions and active days for Claude Code and Codex, by agent, project or model. Every number is read from the transcripts your agents already keep on your machine. Nothing is uploaded, and no server is involved.
Tokens, sessions and streaks
From the last 7 days to a full year, with a daily chart you can split by agent.
A year at a glance
A 12-month activity heatmap and your busiest hours, built from local transcripts and stored only on this device.
What each open tab costs
Memory and tokens for every tab. Close the idle ones in one go, or restart an agent to free its memory and keep the session.
Local usage stats · read from your own transcripts · never leaves your machine
4.0 Resume
Pick up exactly
where you left off
Close a tab, quit the app, or come back tomorrow. Shipwright reattaches to each Claude Code and Codex session, so one click drops you back into the full context.
Restore sessions
Restore all in Rangent⌘⇧R
Markdown tablesThu
Plan putting fixesThu
Wind audio passMon
Per tab or the whole project
Resume a single tab, or bring an entire project's agents back online at once.
History intact
Full conversation and scrollback restored, not a fresh prompt. Survives app restarts.
5.0 Built for long runs
Details that
keep you in flow
Everything you reach for during a long agent run lives in the same window: files, history, diffs, and Sail to keep your screen awake until the run is done.
Built-in file explorer
Browse the project tree beside the active agent without leaving the window.
SailScreen awake
Turn Sail off
Sail for 30 min
Sail for 1 hr✓
Sail for 4 hrs
Sail keeps your screen awake
Click Sail in the sidebar and your screen stays on while agents work, so long runs finish instead of stalling on sleep and you can watch them land. Leave it on, or pick 5 min to 4 hrs and it switches itself off.
On Windows, Sail keeps the system awake; the display can still sleep
5.1
Search your sessions
Find that command, error, or answer from any past run, not just the scrollback in front of you.
5.2
History & autocomplete
Recall what you ran before and complete commands as you type, without leaving the agent.
5.3
Markdown viewer
Plans, specs, and READMEs render as formatted markdown right inside the app.
5.4
Code review panel
See exactly what your agent edited in a built-in diff panel before you accept it.
5.5
Fast native terminal
A GPU-accelerated core keeps scrollback and live output smooth under heavy load.
5.6
Light & dark
A focused dark mode and an airy light mode. The whole interface follows, chrome to agent output.