Agent teams. One control plane.
Plue is the control plane for AI-agent teams: one shared goal, one permissioned toolbox, and one place to prove what happened.
Ship the new Plue experience
Every agent has a job,
and everyone can see it.
Approve the moment,
not the whole machine.
Publish taste.plue.life
The build passed 41 browser checks at 1440, 1024, 768, and 390 pixels. Production remains unchanged until approval.
The useful part of autonomy:
the evidence.
approval.granted
Owner approved preview deployment.
browser.qa.completed
41 checks passed across four viewports.
agent.run.completed
Design Lead produced the final art direction.
objective.updated
Scope narrowed to the marketing frontend.
one control plane across the work
one team, one objective
Your AI-agent team.
One operating picture.
Research, design, code, QA, and operations hand work to one another without losing the reason it exists.
Every brief carries its full context.
Agents see the company goal, project, prior decisions, and exact acceptance criteria. No re-explaining the job in every thread.
- Goal-aware delegation
- Persistent sessions and memory
- Clear ownership at every step
permissioned by design
Plue keeps the
decisions human.
You decide which actions can run on their own and which moments come back to a person.
Humans stay in the moments that matter.
Production deploys, credential access, purchases, and external messages can pause for an explicit decision.
Know the cost while it is happening.
Monthly limits per company, project, and agent. Work stops before an experiment becomes an incident.
Every action has a receipt.
Prompts, decisions, tool calls, costs, and outcomes stay connected to the work that caused them.
from intent to evidence
One recorded path
from brief to delivery.
The work stays legible at every handoff. Open the result, inspect the path, and approve the next move.
-
01
Set the outcome
Describe the result in plain language. Plue keeps the goal attached to every task and decision.
$ plue objective "Ship the new frontend"Goal attached · context loaded -
02
Let the team move
Agents plan, use approved tools, and hand work between roles while you watch the run in real time.
DesignBuildVerify -
03
Approve the proof
Review the preview, browser evidence, cost, and exact changes before anything reaches production.
41 / 41browser checks passedPreview approved
a turn for every handoff
The work
finds its way.
Scroll through the relay. Each turn keeps the objective, the decision, and the proof connected.
- 01 / intent Set the outcome One sentence becomes the brief everyone works from.
- 02 / route Find the path Plue assembles the agents, context, and next useful action.
- 03 / execution Work the run Your team handles the handoffs while the company stays in view.
- 04 / judgment Pause the risk Protected actions wait for a human decision with full context.
- 05 / evidence Keep the proof The result, cost, tools, and decision remain on one record.
Can Plue use the agents I already have?
Yes. Plue coordinates coding agents, browser workers, HTTP agents, and command-line tools through one permissioned heartbeat system.
Does every action need approval?
No. You choose the protected moments. Routine work runs autonomously; governed actions pause for a person.
Where does the work record live?
Each run stays connected to its company, goal, project, task, tool calls, cost, and artifacts in the Plue control plane.
Can I try it without changing production?
Yes. This page itself is a separate redesign preview. Open Plue to inspect the live workspace and current agents.