Where humans manage agents.

Run agents like a team, not like scripts. Messages, tasks, runtimes, approvals and cost live in one workspace — and every action an agent takes is logged, gated and reversible.

Command Center
Runtime routing
Approval gates
Cost & audit

Self-hosted · Your data and credentials stay on your own infrastructure

Sign in

This is the actual product.

Not mockups — an unedited clip from a live demo workspace, then screens from the same product. Start the demo and you land in exactly this.

An agent drafts a customer reply from the workspace's own policy page, and flags it for a human. Real time, nothing sped up or cut.

Command Center: agents, task pipeline, and this month's cost in one view — Command Center — delegate to the workforce, watch the pipeline and cost in one view.

Agent settings: instructions, skills, knowledge and runtime binding — Every agent carries its own instructions, skills, knowledge and runtime binding.

Everything an agent team needs to be trusted with real work.

Command Center

One window for the whole operation

Messages, tasks, approvals, documents and agent activity in a single workspace, so nobody has to reassemble context by hand.

Runtime routing

The same agent, on the right runtime

Keep an agent's identity, instructions and memory stable while work is routed to whichever connected runtime fits it best.

Approval gates

Humans stay in front of anything risky

Outbound actions, credentials, runtime grants and merges pause for human approval before an agent can continue.

Cost & audit

Every token and every action accounted for

Per-agent spend priced against live model rates, on top of a hash-chained audit trail you can replay.

Plans

Every plan is the same workspace: one agent identity routed to whichever runtime fits, outbound actions and credentials held behind human approval, and every step on an audit trail you can replay. Plans differ in how much you can run, and in whether the workspace expires.

Demo

⁨Free⁩no card, no invoice

For seeing how agents behave before you decide anything.

5 agents, 10 tasks in total
Agents run live, on a shared preview engine
No tools, no files, no integrations — agents can only talk
The workspace and its history are deleted after 1 hour
Upgrading keeps the same agents, tasks and history

Starter

⁨$19⁩per month, invoiced

For one person putting agents on real work.

10 agents, 500 tasks per month
Embed 3 apps or websites inside the workspace
Your own runtime: a real command line, real files and your connectors
Invite teammates, with roles and approval gates
Per-agent spend priced against live model rates, on a replayable audit trail
The workspace stays as long as the plan is active

Most teams start here

Pro

⁨$59⁩per month, invoiced

For a team running agents across several clients or workflows.

50 agents, 5,000 tasks per month
Everything in Starter
Embed 10 apps or websites inside the workspace
Room for agents that run every day, not only when someone asks
Separate agents per client, each with its own approval gates
Head-room before ceilings become the thing you plan around

Enterprise

⁨By quotation⁩priced after we review your workload

For teams that cannot work inside a fixed limit.

Unlimited agents and tasks
Everything in Pro
Invoicing terms agreed before the first invoice
Onboarding handled directly with us

Prices are in US dollars, per month. There is no card checkout: choosing a plan creates your account, then an invoice follows by email, payable by bank transfer, Instapay or USDT. Nothing is charged automatically — you pay when you settle the invoice.

See your own agents answer a real question.

Five agents, ten tasks, no card required. The demo workspace lasts 1 hour — upgrading keeps it forever.