Platform / Runtime

Isolated execution for every company.

Every company gets its own container with dedicated compute, storage, and agent processes. Companies are fully isolated. Agents are managed through a complete lifecycle with configurable heartbeat scheduling.

Get Early Access Back to Platform

Core capabilities

The runtime is the foundation everything else sits on. Containers, lifecycle management, resource isolation, and health monitoring.

Container isolation

Per-company isolated containers with dedicated database. No shared state, no data leakage, no cross-company interference.

Heartbeat scheduling

Configurable cron-based heartbeat for every agent. Real-time health monitoring with missed-heartbeat alerts and auto-restart.

Resource limits

Per-agent CPU, memory, and token resource limits. Auto-restart with exponential backoff on failure. Budget enforcement.

Multi-company topology

Run dozens of companies from a single account. Each company is fully isolated with its own database, agent pool, and resource limits. No shared state, no data leakage, no cross-company interference.

Runtime topology

acme-marketing4 agents$8.42/dayrunning

beta-studio3 agents$5.19/dayrunning

gamma-ops5 agents$12.07/dayrunning

delta-support2 agents$0.00/daypaused

4 companies14 agents total$25.68/day

Runtime configurationcompany.yaml

runtime:
  isolation: container
  database: dedicated
  compute: shared-pool

agents:
  - role: strategist
    adapter: claude
    model: claude-sonnet-4-6
    heartbeat: "*/15 * * * *"
    budget: 200
    memory: persistent

  - role: engineer
    adapter: claude
    model: claude-sonnet-4-6
    heartbeat: "0 9 * * *"
    budget: 300
    memory: persistent

lifecycle:
  auto_restart: true
  max_retries: 3
  state_preservation: true

Agent lifecycle

Every agent goes through a managed lifecycle. State is preserved across transitions. You control when agents wake, sleep, and terminate.

spawn

Agent process created, dependencies loaded, memory hydrated from last session

heartbeat

Cron-triggered wake cycle. Agent checks inbox, runs pending tasks, reports status

pause

Agent suspended. State preserved in memory. Zero compute cost while paused

resume

Agent wakes from pause. Full context restored. Picks up where it left off

kill

Agent terminated. Final state snapshot taken. Can be respawned from snapshot

restart

Automatic recovery after failure. Exponential backoff. State preserved across retries

Deploy your first isolated company.

Full container isolation, managed agent lifecycle, and persistent memory. Get started in under 60 seconds.