AutoBots
AutoBots are Autessa's built-in AI copilots, embedded right where you work. Each surface has its own specialist: one designs agents and tools, one manages your database, one writes AutessaScript, one builds entire apps, and one drives live test runs.
What they're for
Everything an AutoBot does, you could do by hand in the same screen — the copilots exist to get you there faster. Describe what you want in plain language and the AutoBot explores what you already have, proposes concrete changes, and applies them once you approve. They are strongest for first drafts and structural work (a new agent with the right tools, a table schema, an app scaffold); you refine the result with the normal editors afterwards.
The five surfaces
Agentic Copilot — lives on the Tools and Agents pages as an Autobot button. It builds agentic workflows from natural language: describe a job and it designs the tools and agents to do it, showing its work through three specialists (Discovery, Builder, Publisher) and proposing creations as change cards you apply or reject. See Agentic Copilot.
DB Copilot — the Autobot button on the AutessaDB Tables page. It manages your database conversationally: create tables, modify schemas, and query data by asking. Schema changes are always previewed — columns, types, foreign keys — before you apply them. See DB Copilot.
Script Copilot — built into every AutessaScript editor (script tools, webhook handlers, computed columns). A chat tab writes and fixes scripts you can apply straight into the editor, and a Docs tab gives you a searchable reference of every helper and built-in. See Script Copilot.
Vibecode — the AutoBot in the App Builder. It generates and edits a complete custom app from chat: it plans first, shows per-file diffs, and refreshes the live preview as changes land. The only copilot with resumable sessions. See Vibecode.
Streaming Test Window — the Autobot (Streaming) panel behind the Test button on agent and multi-agent pages. It doesn't build anything; it runs your agent live, streaming every message, tool call, and (for multi-agents) each sub-agent delegation so you can see exactly how a run unfolds. See Testing.
Comparison
| Surface | Where it lives | What it does | Reach for it when |
|---|---|---|---|
| Agentic Copilot | Tools & Agents pages | Creates tools and agents from a description | You're starting a workflow and want a working first draft |
| DB Copilot | AutessaDB Tables page | Creates tables, modifies schemas, queries data | You know the data you need but not the schema details |
| Script Copilot | Every AutessaScript editor | Writes and fixes scripts in place, with a docs reference | A script tool, validation, or computed column needs code |
| Vibecode | App Builder editor | Generates and edits a full custom app | You want an app built or reworked without writing the code |
| Streaming Test Window | Test button on agents & multi-agents | Runs live tests with full tool-call visibility | You need to see why an agent behaves the way it does |
General guidelines
- Review before applying. Every AutoBot marks its output as AI-generated and (except when you enable Vibecode's auto-apply) proposes changes as cards with explicit Apply Changes / Reject Changes buttons. Read the proposal — column types, tool configurations, file diffs — before accepting.
- Copilots act within your permissions. An AutoBot can only see and change what you can. It won't reach resources your role doesn't grant, and everything it creates is owned like anything you made by hand.
- They consume LLM credits. The Agentic, DB, and Vibecode copilots draw on your company's daily and weekly AI credit allowances; when a limit is reached, the copilot shows a credit-limit banner until the window resets. See Usage & Limits.
- Sessions are mostly ephemeral. The Agentic and DB copilots start fresh each time you open them; only Vibecode keeps resumable conversation sessions.