ApexCorpPay for the orchestration. Pay the model directly.
The subscription is flat and predictable. The variable cost is yours, visible next to the kill switch, and capped by ceilings you set.
One company, the full panel, no card. Enough to find out whether it works for you.
- One company
- 2 agents running at once
- Full control panel and live feed
- Bring your own model key
- No notifications
- No history export
A hosted orchestrator that keeps running when your laptop closes.
- Hosted, always-on orchestrator
- Up to 4 agents running at once
- 90 days of task and note history
- Email on anything that needs you
- Bring your own model key
Several companies, and other people who need to see them.
- Up to 5 companies
- Up to 12 agents running at once
- Unlimited history and export
- Shared access with roles
- Audit log of every approval
Run it in your own environment. We agree terms, then send you everything you need to set it up.
- Setup instructions sent once terms are agreed
- Runs entirely inside your infrastructure
- Concurrency and limits set by your policy
- Named support contact
- Help with security questionnaires
What a company actually costs to run
Model spend dominates, which is why the router is the product’s most important component and why its numbers are on the landing page rather than hidden.
| A one-line fix, including its QA review — measured | $0.0164 |
| The same work before the routing fix | $0.9328 |
| Default daily ceiling, before anything halts | $10.00 |
Deep work — a migration, an audit, an investigation — is classified hard and routed to the top tier, and costs correspondingly more. Under-powering that work costs more in retries than the upgrade does.
Questions worth asking first
Why is model spend not included?
Because reselling tokens would put our incentive against the router. The router exists to spend less — it picks the cheapest model that can do each task, and fixing one classification bug in it cut the cost of a trivial task by 57×. If we marked tokens up, every improvement there would cost us money. You connect your own key and see exactly what the company spends.
What will the models actually cost me?
A measured example: a one-line typo fix, including the QA review it triggers, came to $0.0164. Deep work routed to the top tier costs meaningfully more. The daily and per-run ceilings are yours to set, and crossing one halts every agent rather than continuing quietly.
Why per company and not per seat?
Seats are the wrong unit when the agents do the work — one person can run a company of nine. What scales with the value you get is the number of companies; what scales with our cost is how much of one runs at once. Those are the two things the tiers move along.
Can agents send email, commit code, or post anything?
No. Outward-facing roles get a draft capability and no send capability. No role in the system holds one, and a test asserts that. Everything outward lands in the approval queue for a person.
What happens to my data if I leave?
You can export your company — tasks, notes, memory and the full activity history — at any time while the account is open, and for 30 days after it closes. We do not hold it hostage to keep you subscribed.
Can we run it inside our own network?
Yes, on the Enterprise tier. Once terms are agreed we send you the setup instructions and everything needed to stand it up in your own environment, and stay available while you do. It is not a public download — an autonomous system that runs commands and reads files should be set up by someone who has spoken to us first.
Is the demo the real product?
It is the same control panel, running the same components, against an in-memory company instead of an orchestrator. That is deliberate — a purpose-built showcase would drift from the product within a week.
Start by looking at one
A nine-agent company mid-flight, with an escalation waiting on you. No signup, no key, nothing to spend.