Grex
How it works
Fleet operations for AI agents

Your agents are running.
You have no idea what they're doing.

Grex is the control room for your autonomous AI fleet — real-time visibility, cost tracking, and dead-man alerts across every machine you own.

Running agents at scale is a black box.

You built the agents. You deployed them. Now they're running on three different machines and you're piecing together what happened from log files and gut feeling.

You don't know when something missed a run. You don't know what it cost. You find out something broke when a task is days late.

Your fleet runs while you sleep.

Once configured, your agents run on their own schedule — 24 hours a day, 7 days a week. Researching, building, reviewing, reporting. You check the dashboard in the morning, not at 3am.

Dead-man alerts fire if something stops. Budget windows prevent runaway spend. Every run is logged so you can replay exactly what happened — and exactly what it cost.

  • Agents run on schedule without supervision
  • Missed runs trigger an immediate alert
  • Per-machine budget windows cut off runaway spend
  • Full run history — model, tokens, outcome, cost
  • Jump in when it matters, step away when it doesn't
  • Talk to Hermes from any browser to change config on the fly

You built the agents.
We built the control room.

One dashboard. Every machine. Full visibility from heartbeat to cost.

🖥️

Fleet-wide visibility

Every Nidus (machine), every agent crew, every run — in one view. Know what's healthy, what's drifting, and what's dead on arrival before anyone else does.

dashboard → fleet → nidi → agents
💀

Dead-man heartbeats

If your agent stops checking in, you hear about it. Configurable windows per machine. No more finding out a cron stopped running because a task is three days late.

alert if silent > N minutes
💰

Token burn tracking

See exactly what each agent is spending — per run, per day, per model. Budget thresholds fire alerts before you blow past a limit you didn't know you had.

input · output · cost_usd · model
🚨

Alerts that actually fire

Slack, PagerDuty, webhook — your choice. Configurable rules on error rate, run failures, and missed schedules. Test-fire from the dashboard before you need it.

slack · pagerduty · webhook
📦

Pack-aware fleet view

Agents are organized into Manipulus Packs — self-describing crews with a single manifest. Grex understands your pack structure and shows you exactly what's installed where.

manipulus.yaml → fleet view
💬

Frontier agent chat

Start a multi-turn conversation with Claude, DeepSeek, GPT-4, or Gemini — running on your own hardware, from any browser. No SSH, no terminal, no API keys passed through third-party apps.

dashboard → session → Hermes claims → reply

Your data never leaves your machines.

Every other monitoring tool routes your agent data through their cloud. Grex is different. The dashboard aggregates telemetry from your fleet — but your code, your tickets, your business logic stays where it belongs.

Self-hosted on your infrastructure. Open-core. No vendor lock-in, no surprise data clauses.

  • Agents run on hardware you own and control
  • Heartbeat and run data stays on your infrastructure
  • Framework-agnostic — works with any AI implementation
  • No usage-based data fees or egress costs
  • Full audit log of every operation
  • Open-core: fork it, extend it, own it

Start monitoring your fleet.

Get full visibility into every agent, every machine, every run — today.

Open your dashboard →