Use frontier intelligence selectively
Bring the strongest model when it matters.
A frontier agent can help frame the mission, investigate ambiguity, or review difficult evidence without becoming the system of record for the work.
Autonomous AI workforce on hardware you control
Your computers should work while you're away.
Give Grex a mission. It coordinates AI agents across one laptop, one desktop, or a fleet of machines you control — with the browser, filesystem, compute, and tools needed to finish real work.
Your assistant helps you ask and steer. Grex makes the work durable — coordinating workers, enforcing budgets, checking evidence, and recording an outcome that survives the conversation. Use supported AI subscriptions for volume work instead of making every long-running mission a per-token experiment.
The operating model
Set the mission. Walk away.
Grex turns the computers you already own into an asynchronous AI workforce. Start the work during the day, stay connected from anywhere, and return to outcomes instead of an open chat waiting for its next prompt.
- 01
Give Grex a mission
Define the outcome, evidence, time horizon, economics, budget, and conditions that should stop the work.
- 02
Let the fleet work
Agents plan, delegate, browse, run code, use files, and operate connected tools across your machines.
- 03
Let Grex challenge it
Checks and independent judgments test the artifacts against the mission contract instead of trusting the producing agent's final answer.
- 04
Review the outcome
Return to a ratified result, a decision waiting at an operator gate, or an evidence-backed reason the mission stopped.
Your compute. Your choice.
Start with one machine — scale to as many as the mission needs.
A laptop or desktop you already own is enough. Add another computer when you want more capacity or a different role. Cloud machines are optional, not a prerequisite.
MISSION CONTROLGrex · Your preferred assistantcontract · steer · verify · approve
NIDUS 01Your laptopbrowser · files · compute
NIDUS 02Your desktoprepos · shell · local tools
NIDUS NCloud, if usefuloptional capacity
Execute with budget intelligence
Spend premium reasoning only where it matters.
Budget agents perform the volume work under mission controls. Mix models by role instead of paying frontier-model rates for every action.
Run agents on the subscriptions you already pay for. Grex authenticates against those plans directly, so per-token billing isn't the default path. Bring metered API keys instead if that suits a given mission better.
Built for real work
An AI workforce, not another AI tab
Full access to the computer
Grex agents can use the browser, filesystem, shell, local compute, repositories, and connected services. They work in the same environment where you work.
A fleet, not a bot
One machine can run one or more crews. Add machines as the mission grows. Organize agents around business functions instead of keeping everything trapped in one conversation.
Bring your own intelligence
Use supported subscriptions, metered APIs, or both. Assign different models to different roles and change the underlying engine without changing the mission or its controls.
Autonomy with evidence
Receipts show what changed, budgets bound consumption, and policy gates reserve selected judgments for the operator. Full machine access is paired with an operator-visible record.
A real receipt, from our own fleet
This is a real, ed25519-signed receipt — not a mock. Mission 010 withdrew itself 21 minutes in rather than let a dishonest result settle. Third-party and market names are redacted below; everything else is as issued.
Read the day's full story or see what receipts are for.
What mission would you hand to a fleet of computers working while you are away?
Share your interest