Every instance is a computer
Filesystem, browser, terminal, web search, LLM. Give it credentials and it can do anything you could.
Skills are the unit of work
One skill powers every agent that needs it. Improve once, every agent gets the upgrade.
The manager walks you through it
Chat or API. It asks, suggests, builds, tests, and reports back.
How it works
Build
Describe the task. The manager asks how you will measure success, then creates the first version and wires up integrations.
Test
Multiple instances run different skill variants in parallel. The manager reports results in a comparison table.
Improve
Winners survive, losers get culled. The manager proposes what to try next. You guide it or let it pick. Set up automations to run the loop on a schedule.
Deploy
Winning skill goes to production. Variants get cleaned up. Read more →
Use cases
- GTM
- Recruiting
- Engineering
- Content
- Research
Lead scoring trained on your close data. Pull closed/won and closed/lost deals from Salesforce. Measure against actual conversion rate.Outbound that adapts to reply data. Research prospects via web and LinkedIn. A/B test across batches. Retrain on reply rate each week.Competitor tracking that filters noise. Watch pricing pages, changelogs, and G2 reviews. Post only when something real changes.
What Operator manages
Containers
Provisioning, restarts, storage
AI access
Model access and web search
Variants
Create, score, cull, evolve
Secrets
Encrypted credential storage
Evaluation
Scoring when results come in
Fleet ops
Operations across all instances
Quickstart
Build and test your first skill in one session.
Skills
The build, test, improve lifecycle in detail.
Instances
What an instance can do, types, and statuses.
Automations
Schedule evaluation loops and fleet operations.
Webhooks
Trigger skill runs from external events.
Environment
Store secrets that unlock instance capabilities.
Integrations
GitHub, databases, email, social, and more.
API
Drive everything programmatically.