Pick your engine. Or choose Auto.
Choose an engine for each conversation. Auto routes prompts using the available models and your engine preferences.
For people who build with agents
They code.You lead.
A dashboard for Claude Code, Codex and your other coding agents.
See who's working. Answer who's waiting.
Give the next job a mission.
Across your machines, from your desk or your phone.
Private beta. Bring the CLI accounts you already use.
Your work, in viewAcross your machines
A good morning.3 conversations
Redesign checkout
WorkingFix billing webhooks
Needs youReview auth changes
FinishedOne decision. Then back to work.
How should failed webhooks retry?
Codex found the issue. It needs your call on the retry policy before continuing.
↑ Try it. Unblock the agent.
Illustrative workflow. Your projects, engines and output will vary.
Your agents. Your machines. Your call.
One interface for all your enginesOne interface.
All your engines.
- Claude Code
- Codex
- Grok
- Gemini
- Kimi
- Muse
- Auto
A better way to supervise
You hired the agents.
Then became the tab manager.
One is editing. One is stuck. Another finished ten minutes ago. Finding out shouldn't mean opening every terminal.
Whipcode brings those conversations into one workspace, so you can spend your attention on the work that actually needs you.
01 / Know where you're neededCatch the question.
Catch the question.
Keep things moving.
An agent waiting for an answer is an agent doing nothing. Whipcode brings questions and approvals into one inbox, across your machines.
Read the context. Make the call.
The conversation is right there when you need to dig in.
Live output · Tool calls · Subagent activity · Questions
YOUR ATTENTION2
billing-api / Codex / linux-box
How should failed webhooks retry?Exponential backoffFixed interval
storefront / Claude Code / studio-mac
Keep guest checkout in the new flow?Keep itRequire an account
Different projects. One inbox.
02 / Delegate the follow-throughGive it a goal.
Give it a goal.
Skip the “continue.”
Some work takes more than one turn. A mission lets an agent keep going toward a goal without waiting for your next prompt.
A separate AI judge checks the result and sends it back for another pass when needed. You can inspect progress, pause or stop.
Runs within configured limits. Blocked work still needs you.
MISSION / BILLING-APIUntil done ↻
THE GOAL
Fix webhook retries.
Cover the failure cases with tests.
- Agent implements the fixUpdates retry handling and adds tests.
- Judge finds a missing case“What happens when the retry also fails?”
- Agent takes another passContinues with the judge's feedback.
- You review the resultThe goal and its evaluation stay together.
03 / Step away, stay in the loopYour desk is optional.
Your desk is optional.
Your direction isn't.
Start work on your computer. Check in from your phone. Answer the question that's holding things up, without reopening your laptop.
Your agents keep running on the connected machine. You reach the same conversations through a mobile workspace built for a smaller screen.
Works in your browser. Add it to your home screen.
A quick answer.
Then back
to your day.
A different screen.
Keep your setup. Expand your reach.Your machines.
Your machines.
A much bigger desk.
Connect your laptop, desktop or server to one hub. Whipcode runs the agents where your repos, tools and CLI logins already live.
Your browser becomes the place to see and direct them all.
How connecting a machine works ↓Laptop macOS
Desktop macOS
Server Linux
One workspace.Browser & phone
Client on each machine → your hub → you
Run agents locally
Work happens on the machine with the repo. The hub relays conversations, including the output your agents share.
Connect from behind a router
Clients connect out to the hub. Your agent machines don't need a public IP or an inbound port.
Choose where the hub lives
Use the managed beta or self-host. The core is MIT licensed; source access is by invitation during the beta.
Fits the way you already build.
Your engines, your context, your workflow. Less starting over.
Show the bug, don't just describe it.
Attach a screenshot or a file to give your agent the context it needs, right in the conversation.
Your terminal still has a place.
Resume Claude sessions in its CLI. Open classic tmux terminals from the hub when the job calls for a shell.
Direct more than one conversation.
SuperWhip turns a plain-language instruction into a plan across conversations. Review the plan before applying it.
Before you jump inA few good
A few good
questions.
The practical details, without the fine-print scavenger hunt.
Can I use Claude Code and Codex from my phone?
Yes. Whipcode lets you follow Claude Code and Codex conversations, send instructions and answer questions from a mobile browser. You can manage conversations across connected machines from the same workspace.
The agents run on the Mac or Linux machine you choose, using the CLI authentication already configured there. That machine must stay online and awake. Whipcode is in private beta; request early access to get started.
What do I need to get started?
A Mac or Linux machine with Node.js and the coding CLI you want to use, already signed in. Connect the Whipcode client to your hub, then open the workspace in your browser.
The managed hub is in private beta. Self-hosting is also available with invited access to the repository.
Do I need new subscriptions or API keys?
Whipcode uses the authentication configured for each CLI on your machine. Supported subscription logins work through those CLIs; engine usage remains subject to your provider's plan and limits.
Optional integrations, such as voice, may require their own provider credentials.
Is the agent running in the cloud?
The agent runs on whichever connected machine you choose. That machine must stay online and awake to keep working. Closing the browser doesn't move the job to another computer.
The hub relays conversation traffic. Your agent provider receives the context sent by its CLI, under that provider's terms.
Can I trust a mission to finish on its own?
A mission can continue across turns, with a separate AI judge checking progress against your goal. It can also hit a limit, get blocked or need your decision. You can pause or stop it.
The judge adds another check. It doesn't guarantee correct code or replace testing and your review.
Can I keep using my terminal?
Yes. Claude conversations use the CLI's session files, so you can pick one up with claude --resume <session-id>. Classic tmux sessions are also accessible through the hub.
Session continuity depends on the engine. The Claude resume behavior isn't a promise for every provider.
What about voice, MCP and WhatsApp?
Opt-in voice and the Whipcode MCP server offer other ways to direct your agents. Voice requires provider configuration. MCP lets a connected AI client work with conversations and missions.
WhatsApp commands and alerts require a separately configured bridge on a self-hosted hub.
How much does Whipcode cost?
The self-hosted core is MIT licensed. You cover hosting and your engine providers' subscriptions or usage. Repository access is by invitation during the beta.
Pricing for the managed hub hasn't been announced.
More room to build.
Less to juggle.
You already have the agents.
Give yourself one place to lead them.