What is Ballet?
TL;DR: Ballet is an agent operations platform that helps teams design, run, and maintain repeatable automations — called playbooks — with AI assistance. You describe workflows in plain language, connect your tools, and run them on demand, on a schedule, or via webhooks. Ballet is built by the team behind Brainfish and is currently in early access.
Who this is for
Anyone evaluating or using Ballet - design partners, Builders who author workflows, and Members who run them.
What problem does Ballet solve?
Many teams have an integration backlog: workflows stuck between tools, APIs, and manual steps. Ballet helps you turn those workflows into playbooks — multi-step automations that run reliably with visible progress, logs, and outputs.
How does Ballet work?
Ballet combines four ideas:
- Playbooks — ordered steps (code, HTTP calls, AI agents) that run deterministically.
- Ask Ballet (Console) — an AI assistant that plans, builds, and debugs automations with you.
- Connectors — links to external services via MCP servers and first-party integrations.
- Runs — each execution of a playbook, with step status, history, and shareable results.
What is early access?
Ballet is in an early access / design-partner phase. You request access at ballet.dev/signup, work with the Ballet team on a scoped workflow review, and get design-partner pricing locked in before general availability. General public self-serve signup is not available yet.
What makes Ballet different?
- Reviewable code — steps run in a sandbox; your engineering team can review what the automation does.
- AI-native building — Ask Ballet helps you plan and build playbooks inside the product, not in a separate chat tool.
- Integration-first — connectors, secrets, and MCP servers are first-class, not bolted on.