TL;DR: A quick reference for Ballet terms you'll encounter in the Studio, help articles, and Ask Ballet conversations. Terms are grouped by category — platform, automation, AI, and admin.
Who this is for
All Ballet users looking up unfamiliar terms.
Platform terms
| Term | Definition |
|---|
| Workspace | Your team's shared Ballet environment (company/account). |
| Owner | Role with billing, invites, and full builder access. |
| Builder | Role that creates and edits playbooks, connectors, and agents. |
| Member | Role that runs granted playbooks; cannot author. |
| Builder Mode | Full Studio UI for authoring automations. |
| Member Mode | Simplified UI for running playbooks and chatting. |
| Studio | The Ballet web application (app.ballet.dev). |
Automation terms
| Term | Definition |
|---|
| Playbook | A multi-step automation workflow. |
| Step | One unit in a playbook (Code or HTTP). |
| Run | One execution of a playbook. |
| Trigger | What starts a run: Manual, Schedule, or Webhook. |
| Collection | A folder organizing playbooks; may be Git-backed. |
| Input | A value the runner provides when starting a playbook (flow.input). |
| Output | Data a step or run produces. |
| Suspended / Paused | Non-terminal run status while waiting to continue; sandbox is freed; same runId. |
| Pause | Operator action that parks a run between steps after the current step finishes. |
| Continue | Wake a suspended run on the same runId (Studio or API). |
| waitForInput / suspend | SDK call that parks mid-step for human or external signal (reason: input_request). |
| state.resume | Park payload on a suspended run: requestId, stepId, prompt, details, schema, reason. |
| requestId | Server-issued correlation token required to wake a HITL park via /reply. |
| Drift | Git sync state when remote and Ballet versions differ. |
AI terms
| Term | Definition |
|---|
| Ask Ballet / Console | The AI assistant panel on Home and inside playbooks. |
| Agent | A configured AI assistant with model, instructions, tools, and skills. |
| Skill | Reusable AI instruction template for specific task types. |
| Tool | A callable function used by agents or playbooks. |
| Plan / Build | Console flow: AI proposes a plan, then executes on confirmation. |
| Allow / Ask | Console permission modes for autonomous vs. approved actions. |
Connection terms
| Term | Definition |
|---|
| Connector | A linked external service or MCP server. |
| MCP | Model Context Protocol — standard for connecting AI to tools. |
| Secret | An encrypted credential stored in the workspace. |
| AI Gateway | Workspace setting for AI model routing and providers. |
Related articles
