Join at the speed of context.
Walk into a task with its objective, decisions, changes, and next steps already in view. Less replaying the conversation. More contributing to it.
Multiplayer, by design
Great minds.
Shared context.
Your team and your coding agents, in one workspace. Pick up the context, shape the work, and build what comes next. Together.
Independent minds. Connected work.
A new way to build with AIBuilt by Retia Labs
One task. Everyone in the loop.
Step inside a shared agent workflow.
Storefront / Shared tasks
In progressCheckout recovery
Maya joined the task
You’re up to speed.Catch-up
Alex started checkout recovery. Codex has traced the failure path. The retry rules still need a decision.
The conversation, decisions, and source context travel together.
Illustrative workflow. Explore how a task stays shared.
Bring the tools you think with.
Codex First upClaude Code PlannedCursor Planned
Less catching up. More moving forward.
Built for the space
between great ideas.
The context, decisions, and people behind the code.
Finally in the same place as the agent doing the work.
A better idea can come from anyone.
Give your agent a new direction from inside the shared task. See who sent each instruction and whether it reached the agent.
Pass the work. Keep the thread.
Hand responsibility to a teammate with a note and the full context attached. The task stays on its original execution host.
The right human, in the loop.
Route a pending action to a teammate with approval rights. One clear decision, tied to the exact action and the person behind it.
A shared rhythm
From your first thought
to your team’s next move.
Start a coding task on your machine with your supported Codex account. Make progress on your own, then invite your team when you’re ready.
A useful workspace, even before the first invite.An invited teammate joins from desktop or web, catches up on the work, and adds the perspective the task was missing.
Shared context. Attributed contributions.Review the resulting changes, request help, or hand responsibility to someone else. The work carries its history into every next step.
Your issues and pull requests stay part of the story.The best ideas don’t happen alone. 
Your AI workflow shouldn’t either.
Closer together. Still in control.
Your workspace.
Your boundaries.
Collaboration should open up possibilities.
You decide what it opens up access to.
Private, from the start.
Plexus is being built around private teams, explicitly shared projects, and end-to-end encrypted collaboration. Your chosen AI provider still receives the inputs it needs to do the work.
Your machine. Your agent.
Tasks run on a connected execution host using your supported provider account. Bring teammates into the work while keeping the host and credential ownership clear.
Participation with intention.
Invite named teammates, share the right projects, and delegate approval rights separately. Joining a task doesn’t automatically grant permission to approve its actions.
Join the early-access list
Your next build
is a team sport.
We’re opening Plexus to small teams who want a better way to build with AI. Join the list and help shape what comes next.
Early days. Shared ambition.
Make room for your team.
Join the list for a first look at Plexus.
How we use your details
Retia Labs uses your email address and any optional details to contact you about Plexus early access. Signup details are stored with Brevo. You can request deletion by emailing [email protected].
A little more context.
What is Plexus?
Plexus is a multiplayer workspace for coding agents. It brings your team into shared agent tasks to catch up, give instructions, review changes, and hand off responsibility. It fits around your codebase and existing issues and pull requests.
Can I use it yet?
Plexus is in development. We’re collecting interest for invite-only early access and will contact teams as places become available. Joining the list does not require payment or commit you to a plan.
Which coding agents will it support?
We’re starting with Codex. Claude Code and Cursor are planned as separate follow-on releases. Each provider will have its own verified capabilities. You’ll use your own supported provider account; provider usage is separate from Plexus.
Does everyone need the desktop app?
The plan is desktop execution clients for macOS and Windows, with browser access for invited collaborators. The execution host needs to stay connected and awake. Handing off responsibility doesn’t move the running task to another machine.
How are you approaching privacy?
End-to-end encrypted collaboration and customer-controlled recovery are requirements for early access. Only authorized endpoints should be able to read shared task content. Your chosen inference provider still receives task inputs, and authorized devices and delivered client software remain part of the trust model.