Agent runtime
Internal agents use Auto, Fast, or Deep modes. Deep work can go on for days or weeks: every step is checkpointed, and the next run picks up where the last one stopped.
Agent runtime
Internal agents use Auto, Fast, or Deep modes. Deep work can go on for days or weeks: every step is checkpointed, and the next run picks up where the last one stopped.
Persistent sandboxes
Isolated containers with browser, terminal, and filesystem — one session per agent, reused across runs.
Multi-modal content
Generate images and videos (plan-dependent), plus presentations, diagrams, and documents.
Event-driven Loops
Trigger agents from GitHub, Slack, Jira, inbound email, Telegram, WhatsApp, and cron schedules.
Skills
Reusable workflow templates you attach to agents — invoked by the agent when relevant.
Built-in capabilities
A sandbox, browser automation, web search, Knowledge retrieval, and the Integrations in the catalog.
Define your agent
Describe what your agent should do and how it should behave.
Connect Integrations and Knowledge
Link GitHub, Slack, your documentation, and any service with an MCP endpoint.
Deploy anywhere
Web widget, Telegram, WhatsApp, Slack (internal agents), or REST API.
Monitor and improve
Each run’s steps, results and cost show on the Runs page and in the agent’s Activity. Refine the instructions from what you see.
Quickstart (10 min)
Create and run your first AI agent in under 10 minutes. Start now →
Explore use cases
See real-world examples and templates for common scenarios. Browse →
Read the docs
Guides for agents, Integrations, and deployment. Learn →
Get help
Email support, or report a problem from inside the app. Support →
Read these three pages in order: