A quiet workspace for coding agents

Manage decisions, not agents.

Ruddr runs your coding agents in silence and reaches you only when something waits on you: a decision to take, a design to approve, a blocker to lift.

7 days free, then from $7.99 a month

Native on
  • macOS
  • Windows (WSL)
  • Linux
  • Android
  • iOS
Ruddr's window: a decision waits in Needs you, its card in the chat and its options in the side panelRuddr's window: a decision waits in Needs you, its card in the chat and its options in the side panel
The same decision on the phone, with Choose AThe same decision on the phone, with Choose A
01 · Your agents

A light layer over your real CLIs.

Ruddr is a lightweight layer on top of Claude Code and Codex, as you installed them. Built on herdr-core, your agents run in the background in real terminals, signed in as you, and you can open each one directly. Close the app and they keep working; open it again and everything is where you left it.

No lock-in, and nothing left behind: your sessions stay theirs and resume anywhere, with Ruddr or without it.

Claude Code's own screen in its terminal, opened from its chat in RuddrClaude Code's own screen in its terminal, opened from its chat in Ruddr
02 · Ask

Start with a message.

We like chats for a reason: they're a convenient entry door, and flexible in what you put through it. Here, the agent's tool calls fold into lines, its words stay readable, and how long it has been at work is in sight.

The same session as a chat: your message, the agent's lines folded, its words, and how long it has workedThe same session as a chat: your message, the agent's lines folded, its words, and how long it has worked
03 · Silence

“Which of these notifications actually matter?”

Then, silence.

While agents work, nothing calls you until the task is done: not a step along the way, not a handoff between agents, not a passing test. Needs you holds only what blocks an agent, so when it's empty, nothing needs you, and nothing says so.

92% fewer pings while agents work
Agents at work in several chats, and no Needs you: nothing waits on youAgents at work in several chats, and no Needs you: nothing waits on you
04 · Pages

See it before it's built.

Designs, plans and reports come as pages. Pin a note on anything you see; the agent answers in its thread, and its next version answers the note. Once a page is settled, it gets built as drawn: what's left to say is what you change your mind about.

37% fewer tokens for each note on a design
A design page with a note on one of its rules, the agent's reply, and the version that answered itA design page with a note on one of its rules, the agent's reply, and the version that answered it
05 · Decide

“…unstructured messages that we need to read and read again.”

A decision, when it's yours.

A question comes as a decision: its options side by side, with their captures and the agent's recommendation. Answer with a click, at your desk or on your phone, and the work goes on. Settled, it links the commits that built it.

85% fewer words in the chat: choices come as decisions
A decision waits on you: its card in the chat, its options and their captures in the side panelA decision waits on you: its card in the chat, its options and their captures in the side panel
06 · Team

A chat that becomes a team.

When the work needs more hands, the chat becomes a team: an implementer builds, a reviewer reviews each round, and you keep talking to the same chat. Claude Code, Codex, or both: with both, each reviews the other's work.

From the third round, a finding that repeats an earlier fix stops the work and asks you, before it spends more of your plan.

The chat's team: an implementer at work, a reviewer between rounds, a prototyper done, each with its agentThe chat's team: an implementer at work, a reviewer between rounds, a prototyper done, each with its agent
07 · Queue

“How do you track the progress on a chat? Well, you have to ask.”

Ask and forget. Your agents won't.

Ask for the next thing while the team is still on this one. Each ask joins a member's queue as a unit, kept by Ruddr rather than in an agent's context, so a compaction or a restart forgets none of it. Today and Next draws the day: a lane for each agent, each unit as it's built, reviewed and fixed, and what's queued after now.

An agent on its own keeps the same as a list: in progress, up next, and done until you clear it. Looking costs your agents no tokens.

Today and Next for Search's team: a lane for each agent with its units' bars, a wait on you hatched, and the implementer's queue after now, its next unit's card openToday and Next for Search's team: a lane for each agent with its units' bars, a wait on you hatched, and the implementer's queue after now, its next unit's card open
08 · Ready

Told when it's ready.

You hear from it at the end, when the work is ready for you, or when something blocks it, and not in between. Try it, then review it where it was done: the branch beside its chat, your notes on its lines.

A check in the chat: try search, then review its branch, with what the team did folded above itA check in the chat: try search, then review its branch, with what the team did folded above it
09 · Review

Close to the code.

Like a terminal, and unlike a bot, Ruddr keeps you close to what your agents write. Open a chat's branch or a pull request as a review, beside its chat: its files and commits, the code highlighted, each definition a click away. A note on a line goes to the agent that wrote it, which answers in the note's thread.

A project's Code holds all its work to review, and the repository's files too, each read with its language server and open to a note on any line.

A review of Search's branch: its files by where they stand, the code highlighted, and your note on a line with the agent's answerA review of Search's branch: its files by where they stand, the code highlighted, and your note on a line with the agent's answer
10 · Chats

Quick, or for keeps.

A chat can be as quick as a terminal or as lasting as a bot. Open one with no role for a question, and archive it once it's answered, as you'd close a tab. Keep a coordinator's for a whole area of work, and come back to it as you would to a bot: its state, its pages and its team are where you left them.

Claude Code or Codex, on the model and effort you choose; with a role, its own instructions.

A new chat with its chips, its project, No role and Opus · high, and its role menu open: Coordinator, No role, Implementer, Prototyper, ReviewerA new chat with its chips, its project, No role and Opus · high, and its role menu open: Coordinator, No role, Implementer, Prototyper, Reviewer
11 · Projects

Every project, in its place.

Each project keeps its chats and their teams, its decisions, pages and notes, and its Code, the work there is to review. Its overview shows every agent and where each one stands; archived chats wait at its foot, a click from coming back.

Your projects sit at the rail's foot, with a dot on any that has something for you.

atlas's overview: Search's team, done, Onboarding at work, and two archived chats; the projects at the rail's footatlas's overview: Search's team, done, Onboarding at work, and two archived chats; the projects at the rail's foot
12 · Anywhere

All of it, on your phone.

Not a companion app: everything you do at your desk, you can do on your phone, made for a small screen. Chats, decisions with their captures, pages with their notes, teams and reviews. A push says something waits, never what.

macOS, Windows with WSL and Linux at your desk; Android and iOS in your pocket.

Ruddr on the phone: Needs you, then every chatRuddr on the phone: Needs you, then every chat
A decision on the phone, with Choose AA decision on the phone, with Choose A
The chat on the phone, its settled decision in itThe chat on the phone, its settled decision in it

How much do you want peace?

Peace lives in the smallest details.

  • Nothing lost.Your computer crashes, your connection drops, your limit runs out: each agent comes back where it was, on its model and effort, and by itself once the connection or the limit returns.
  • Close it, update it.Quit Ruddr or update it, and your agents keep working: they run apart from the app. A new Claude Code or Codex reaches each one at its next rest.
  • Step away.Leave your desk, and what matters still finds you, on your phone.
  • Only what matters.Nothing else reaches you: no step along the way, no handoff between agents, no passing test.
  • A clean start.Begin the next piece of work in a new chat. The last one keeps its state, its pages and its team where you left them, with no terminal or tab of it to close.
  • Nothing to hold in your head.Something to say while the agent works? Don't wait: send it in a thread, or let the agent shape it into a page or a decision. Not ready yet? Write it as a note on the project, and take it to an agent when it is.
And the rest of the room

Everything else you'd look for.

Set a message aside

Halfway through a message when the agent turns to something else? Set it aside with a right-click on Send, steer the agent, then bring your message back. It waits on your phone too.

Threads in any chat

Split a long chat by topic. Each thread closes when it's done, so nothing is left half-answered.

Your app, beside its chat

Open what your agent serves as a tab in its chat. It stays as you left it while you work elsewhere, so a test you're halfway through waits for you.

Bring your agents in

Ruddr finds the Claude Code and Codex sessions already running in your repositories and brings them in, each with a name and a face.

Roles and templates

Give a role its instructions, model and effort, and save a running team as a template: the next one starts from it, in its project or in all of them.

Context and limits in sight

Each chat shows its context, and resting agents compact before their cache goes cold. The bar shows CPU, real RAM, disk and what's left of your week, and Pause gives an agent's memory back, about 480 MB each.

Yours, as they are

Native. Local. Nothing to undo.

Native on every screen

Compiled for each platform, not a web page in a wrapper: macOS, Windows with WSL, Linux, Android and iOS, the same app at your desk and in your pocket.

Your machine only

Agents, code and transcripts stay on your machine, and Ruddr keeps no copy of your chats. Its cloud knows your account and your plan, and passes your phone sealed pushes.

Nothing to undo

Ruddr's hooks come with the sessions it launches, never in your settings. Every skill, plugin and hook your agents have works as it does today.

Where Ruddr sits

Between terminals and bots, a place to decide.

Each level is right for some work. If yours is in between, Ruddr is made for it.

Worktree and terminal managers

Terminals, tabs and worktrees, side by side

You manage repositories, branches, worktrees and terminals, every agent in sight. Right when you want your hands on each one.

Ruddr

A workspace built around decisions

Agents work in silence, as many as the work needs, and reach you with what takes a person: a decision, a design to approve, a blocker, work to review where it was done. You keep the final say on what ships. Right when the work goes to other people: a team, an open-source project, a carefully crafted product.

Autonomous bots

Bots that run like a company

Named agents trade work in group chats and carry it through on their own. Right for routines and operations that run without you.

Marcelo Soares
Who makes it

Built with Serverpod.

I'm Marcelo Soares, Serverpod's dev lead, and Ruddr is Flutter on every screen and Serverpod behind it, on your machine. As I ran more projects with agents, my attention spread across spinners, tabs and worktrees, instead of feeling free until an agent really finished. I wanted to put agents to work and hear from them when the work is done or a decision is needed, so I built Ruddr around decisions. Everyone has the same agents now; what sets the work apart is knowing what to build, and how. Ruddr's how is the one that built it, since its fourth day: designs settled on pages, choices as decisions, every change reviewed before it lands.

Pricing

Seven days free. Then one of two plans.

Desk

Ruddr on your computer.

$7.99a month
$79a year
  • All of Ruddr, on the computer your agents run on
Start 7 days free Get early access

Anywhere

Recommended

Ruddr on your computers and your phones.

$9.99a month
$99a year
  • Everything in Desk
  • The phone app: all of Ruddr, on Android and iOS
  • Pushes for what waits on you, and nothing else
  • Your other machines, in the same app
Start 7 days free Get early access

Your card comes first, and nothing is charged until day 7: cancel before, and you pay nothing. A year costs ten months. Model use isn't included: your agents run on your own Claude and Codex plans. Prices in USD, with tax as your country applies it. Sold by Paddle, our merchant of record.

Questions

Do I need a Claude or Codex plan?

Yes. Ruddr runs Claude Code and Codex signed in with your own accounts, so your agents use the plans you already have. One is enough; with both, each can review the other's work. Other agents come next: tell us which you use when you join.

Will my agents behave differently?

They're the same agents. Ruddr launches them with its hooks and a small skill, so they can reach you through decisions instead of paragraphs. Your own settings stay as they are.

Where does my code go?

Nowhere new. Agents, code and transcripts stay on your machine, and Ruddr keeps no copy of your chats. Its cloud holds your account and your plan, and nothing else of yours.

Can I still see the terminals?

Whenever you want. Each agent runs in a real terminal you can open from its chat. Ruddr just doesn't make you watch it.

Which computers and phones does it run on?

macOS, Windows with WSL and Linux at your desk; Android and iOS on your phone. The phone isn't a companion: everything you do at your desk, you can do there, made for a small screen.

Is it a native app?

Yes. Ruddr is compiled for each platform with Flutter, not a web page in a wrapper: its lists, threads and motion are drawn by the app itself, at your desk and on your phone alike.

What happens if I stop paying?

Ruddr turns read-only: every chat, decision and page stays open to read. Your agents keep running in their terminals, and Ruddr shows you the command that reaches them.

Is there a plan for teams?

Soon: team plans are on their way. Until then, a plan is one person's, on all their machines and phones.

Where do the numbers come from?

From Ruddr's own build. It has built itself since its fourth day: in its first 14 days, 436 features and fixes landed, a commit each, 268 of them through Ruddr. Against the same agents run raw:

  • 92% fewer pings while agents work. Run raw, the agents stopped 14.6 times for each feature or fix, each a “done” to check; through Ruddr, 1.2.
  • 95% less to read between your messages. Woken by a teammate, a note or a finished job, your agent answers where it was asked and ends its turn with a lone ·, as 95% of those turns now do: about 30 bytes a turn, against 633 run raw.
  • 85% fewer words in the chat, for the same work. 4.0 KB for each landed feature run raw, 0.6 KB through Ruddr: choices come as decisions and designs as pages, each read once.
  • 37% fewer tokens for each note on a design. A page's notes go to the agent together, where an artifact's comments are read one at a time: 0.17M weighted tokens a note, against 0.27M.
  • About 480 MB freed by each paused agent. 370 to 876 MB across 12 Claude agents.

One team's build, ours, in November 2026: your numbers will differ.

More room. Less noise.

Remember moving into a new office, with more space, more quiet and room to work? Welcome to Ruddr.

Also for Linux and Windows with WSL. Your phone pairs from the desktop app.