Kai - AI Assistant
What Kai can do in your project, why to use it, how it differs from the Keboola MCP Server, how to get started, and where to find security and support details.
Kai is Keboola’s embedded AI assistant—a context-aware data engineering co-pilot that understands your project’s transformations, flows, table schemas, and jobs.
What Kai Can Do
Section titled “What Kai Can Do”Analytics & Exploration — Query databases, explore schemas, calculate metrics, and analyze trends.
Integration Setup — Configure extractors, destinations, and custom components. Manage authentication and packages.
Coding — Write and optimize SQL/Python transformations. Convert between languages.
Troubleshooting — Debug job failures, resolve configuration errors, and investigate data quality issues.
Documentation — Generate project documentation, update table descriptions, and create onboarding materials.
Data Modeling — Build analytical frameworks, dimensional models, and complex data structures.
Semantic Layer - Build and maintain your project’s semantic layer so every AI assistant shares the same definitions of your metrics and business terms.
Why Use Kai?
Section titled “Why Use Kai?”Context-aware — Unlike generic AI tools, Kai reads your actual job logs, configurations, and data structures to provide specific solutions.
No setup required — Built into Keboola, already authenticated, understands your current location.
Data engineering focus — Deep knowledge of Keboola components, transformations, and best practices.
Kai vs. MCP Server
Section titled “Kai vs. MCP Server”| Feature | Kai | MCP Server |
|---|---|---|
| Best for | Browser workflows, team collaboration | IDE integration, local development |
| Setup | None | Requires external tool configuration |
| Audit trail | Full organizational visibility | Local only |
Getting Started
Section titled “Getting Started”Every user on a supported stack can see the Kai Agent button in the project’s top bar, but the feature has to be switched on first:
- Organization Admins can enable Kai directly from the chat screen
- Other users can request access from their Organization Admin or contact Keboola Support
- You can also enable Kai via Settings → Features
Learn More
Section titled “Learn More”- Settings (Tool Permissions & System Instructions)
- Use Cases & Examples
- Best Practices
- Python Client
- Semantic Layer
Security & Privacy
Section titled “Security & Privacy”Your data is never used for AI training and is automatically deleted after processing.
Learn more about Kai security →
Support
Section titled “Support”Use the thumbs up/down buttons in the chat to provide feedback, or contact Keboola Support for access requests and technical issues.
Was this page helpful?
Thank you for helping improve Keboola's documentation!