Tutorial: Discover and use Unity Catalog Skills
Connect your coding agent to the skills your team published to Unity Catalog, then discover and use them from any agent session. Everyone points at the same governed schemas, so you reuse what exists instead of rebuilding it. To author and publish your own skills, see Create and share Unity Catalog Skills.
You use a skill one of two ways: download it to disk so it loads every session, or load it live from the Databricks Skills MCP server to always get the latest published version.
Requirements
- A Unity Catalog-enabled Databricks workspace, and your workspace URL (for example,
https://my-company.cloud.databricks.com). - Python 3.12+ and uv on your machine (used to install
ucode). - An MCP-capable coding agent (MCP, the Model Context Protocol, is the open standard agents use to connect to tools).
USE CATALOG,USE SCHEMA, andREAD VOLUMEon the skills you want to use. Your Databricks admin grants these. If a command fails with a permission error, ask your admin for these on your team's schema.
Connect to the Databricks skill registry
Install ucode and use it to connect your coding agent to your Databricks workspace's skill registry.
ucode signs you in to your workspace, and registers the Databricks Skills MCP server (databricks-skill-registry) that your agent uses to list, load, and download skills.
Have your agent install and connect ucode for you:
Install ucode from its Git source and connect my coding agent to Databricks:
1. Run: uv tool install git+https://github.com/databricks/ucode
2. Run: ucode configure --agents <your-agent> --workspaces https://<workspace-host>
3. Run: ucode configure skills
Use <your-agent> = my coding agent (claude, codex, gemini, opencode, or copilot) and
<workspace-host> = my workspace URL host (for example, my-company.cloud.databricks.com).
A browser will open during step 2 for me to sign in.
To do it yourself, run the same commands in your terminal:
uv tool install git+https://github.com/databricks/ucode
ucode configure --agents <your-agent> --workspaces https://<workspace-host>
ucode configure skills
Replace <your-agent> with the name of the coding agent you use (for example, claude, codex, gemini, opencode, or copilot), and <workspace-host> with your workspace URL host. A browser opens during ucode configure for you to sign in. Restart your agent afterward (ucode <your-agent>) so it loads the new tools.
Discover skills
Ask your agent in plain language; it calls the matching tool on the databricks-skill-registry MCP server. You only ever see skills you have permission to use.
List a schema to see every skill in it:
List the Databricks skills in acme.sql_skills.
The agent calls the list_skills MCP tool and reports what the schema holds.
Use skills
There are three ways to use a skill you found:
- Load a skill once: for this session only. Nothing is saved. Best for a one-off.
- Download a skill: for every session. Downloads to disk so it loads automatically each session, and works offline. Best for the handful you use daily. Copies are point-in-time; re-download for updates.
- Load a schema live: always the latest. Scopes the MCP server to a whole schema so every skill in it stays available and current, with nothing stored locally. Best for a big, fast-changing set, but needs a connection.
In each case, the prompt is what you ask your agent, followed by the tool it calls or the command it runs.
Load a skill once
Use acme.sql_skills.databricks-sql-guide to review this query.
The agent calls the load_skill MCP tool to load the skill's instructions into context, and get_skill_files if the skill bundles supporting files.
Download a skill
Download acme.sql_skills.databricks-sql-guide with ucode so I have it every session:
run ucode configure skills --location acme.sql_skills --skill databricks-sql-guide.
ucode configure skills --location acme.sql_skills --skill databricks-sql-guide # one skill
ucode configure skills --location acme.sql_skills --skill databricks-sql-guide,dbt-conventions # several skills
ucode configure skills --location acme.sql_skills # a whole schema
--location is the schema that holds the skills; --skill narrows the download to a comma-separated list of skill names within it. Omit --skill to download every skill in the schema.
Restart your agent afterward (ucode <your-agent>). Downloaded skills live in your local skills folders (.claude/skills/ and .agents/skills/) and load every session. Try one without naming a command; the agent matches the task to the skill and follows it:
Using our SQL guide skill, write a query on samples.nyctaxi.trips for average fare by trip distance.
samples.nyctaxi.trips exists in every workspace, so this works with no setup.
A downloaded skill is a point-in-time copy. Re-run the download for a newer version, or load the schema live (below) to always get the latest.
Load a schema live
Scopes the Databricks Skills MCP server to a schema so every skill in it is exposed as a live tool.
Load the skills in acme.sql_skills live so I always get the latest version:
run ucode configure skills --location acme.sql_skills --mcp.
ucode configure skills --location acme.sql_skills --mcp
Restart your agent afterward (ucode <your-agent>) so the schema's skills load. Every skill in acme.sql_skills then appears as a live tool named skill_acme.sql_skills.<skill>. When a task matches a skill's description, the agent loads it on demand; publish a new version and everyone on the connection gets it immediately. This pairs well with a schema that a central team keeps in sync from Git.
Live loading works on a whole schema at a time. You can't scope it to a single skill. To use just one skill without loading its whole schema, load it for a one-off instead.
For supported agents, ucode configures the connection to load live skills automatically. Other agents may expose their own "always load" (or equivalent) setting that you'll need to turn on.
Reference
ucode is open source. For the full, current command reference, see the ucode repository. The commands this tutorial uses appear inline in the steps above.
Skills tools (your agent calls these on the databricks-skill-registry MCP server once connected):
Tool | Purpose |
|---|---|
| List the skills in one schema |
| Load a skill's instructions into the current session |
| Fetch specific bundled files from a skill |
Next steps
- What are Unity Catalog Skills for the concepts behind this tutorial.
- Create and share Unity Catalog Skills to publish your own.
- Govern skills for the privileges that control who can read and share skills.