databricks-agent-bricks

Create Databricks Knowledge Assistants and Supervisor Agents via the Databricks CLI.

4|1|Updated May 22, 2026

Search Tags:#rag #multi-agent orchestration #databricks #Knowledge Assistant #supervisor agent #databricks cli #agent bricks

Ready to run

npx skills add https://github.com/ThomazRossito/ai-data-agents --skill databricks-agent-bricks-thomazrossito

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: databricks-agent-bricks
Source: https://github.com/ThomazRossito/ai-data-agents/tree/main/plugins/ai-data-agents/skills/databricks-agent-bricks
Command: npx skills add https://github.com/ThomazRossito/ai-data-agents --skill databricks-agent-bricks-thomazrossito

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Setting up Databricks Agent Bricks requires knowing the exact CLI command shapes, JSON payloads, and resource path conventions for Knowledge Assistants and Supervisor Agents, which are easy to get wrong. ## Core Features & Use Cases - Knowledge Assistants: Create document Q&A agents with RAG over PDFs, TXT, MD, and DOCX files in Unity Catalog Volumes, or backed by an existing Vector Search index. - Supervisor Agents: Build multi-agent orchestrators that route queries to Genie spaces, Knowledge Assistants, UC functions, external MCP servers, and other tool types. - Use Case: A data team creates a Knowledge Assistant over product documentation in a Volume, then attaches it as a tool to a Supervisor Agent alongside a Genie space, so users get both document answers and SQL analytics through one endpoint. ## Quick Start Use the databricks-agent-bricks skill to create a Knowledge Assistant over the files in my Unity Catalog Volume and attach it to a new Supervisor Agent.

Frequently Asked Questions about databricks-agent-bricks

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I create a Knowledge Assistant in Databricks?▼

Run databricks knowledge-assistants create-knowledge-assistant with a name and description, then add a knowledge source with create-knowledge-source passing the parent path positionally and display_name, source_type, and the files or index body inside --json. Finally sync with sync-knowledge-sources.

How do I build a multi-agent Supervisor Agent in Databricks?▼

Use databricks supervisor-agents create-supervisor-agent with a display name plus --description and --instructions flags, then attach tools with create-tool. Each tool needs a tool_type such as genie_space, knowledge_assistant, or uc_function with its type-specific block in --json.

What file formats does a Databricks Knowledge Assistant support?▼

Knowledge Assistants support PDF, TXT, MD, and DOCX files placed in a Unity Catalog Volume. Alternatively, you can point the source at an existing Vector Search index by specifying index_name, text_col, and doc_uri_col.

Why is my Knowledge Assistant stuck in CREATING status?▼

A Knowledge Assistant can take up to 10 minutes to leave CREATING status. If it persists, check workspace quotas, verify the Volume path exists including the trailing slash, and confirm file permissions and supported formats.

Can a Databricks Supervisor Agent call external MCP servers?▼

Yes, attach an external MCP server as a uc_connection tool backed by a UC HTTP Connection created with is_mcp_connection set to true. Store the client_secret in a Databricks secret scope and grant USE CONNECTION to the agent service principal.

Why does my Supervisor Agent endpoint not answer queries after creation?▼

The serving endpoint takes up to about 10 minutes to come online after create-supervisor-agent, even though get-supervisor-agent returns the endpoint name immediately. Check readiness with databricks serving-endpoints get before querying.