add-tools

Add MCP servers and grant Databricks permissions in databricks.yml.

4|Updated May 9, 2026
One-click install
npx skills add https://github.com/victorlou/housing-assistant --skill add-tools-victorlou
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: add-tools
Source: https://github.com/victorlou/housing-assistant/tree/main/app/app-templates/.claude/skills/add-tools-langgraph
Command: npx skills add https://github.com/victorlou/housing-assistant --skill add-tools-victorlou

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill resolves agent tool configuration and Databricks permission errors by guiding you to add MCP tools and grant the required access in databricks.yml so the agent can run them at runtime.

Core Features & Use Cases

  • Add MCP server tools to your LangGraph agent: Connect MCP servers (including Genie spaces, vector search, UC functions, or custom MCP apps) and load their tool definitions into the agent.
  • Grant required access in databricks.yml: Adds the correct resources.apps.<app>.resources entries so deployed bundles have the needed permissions (e.g., CAN_RUN, CAN_USE, CAN_QUERY, EXECUTE, or USE_CONNECTION depending on resource type).
  • Prevent common deployment/runtime failures: Ensures value_from references match resource name fields and recommends MCP error-handling patterns (handle_tool_error, timeout adjustments) to avoid agent crashes.

Quick Start

Use the add-tools skill when you get a runtime permission error or you want to connect a new MCP server or Genie space by updating agent_server/agent.py and adding the matching resources permission entries in databricks.yml, then redeploy and rerun the bundle.

Frequently Asked Questions about add-tools

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

FAQPage Schema
How do I fix Databricks permission errors when running a LangGraph agent?

To fix Databricks permission errors for a LangGraph agent, you add the matching resource permission entries in databricks.yml and wire the MCP tools correctly in agent_server/agent.py. You then redeploy the bundle and rerun the agent entry.

How do I add a Genie space or Unity Catalog function as an MCP tool to my agent?

You add a Genie space or Unity Catalog function by loading the corresponding MCP server tool definitions into your agent and granting the required access in databricks.yml. This ensures the deployed bundle has the necessary permissions like CAN_QUERY or EXECUTE.

What permissions do I need to configure in databricks.yml for MCP tools?

Databricks.yml requires permissions like CAN_RUN, CAN_USE, CAN_QUERY, EXECUTE, or USE_CONNECTION depending on the resource type. You add these as resources.apps.<app>.resources entries so the deployed bundle has the required access for the MCP tools.

Why does my agent crash when calling an MCP tool at runtime?

Agent crashes during MCP tool calls often happen when value_from references do not match resource name fields. You can prevent these failures by implementing MCP error-handling patterns like handle_tool_error and adjusting timeout settings in your code.

Do I need to redeploy my bundle after adding new MCP servers to the agent?

Yes, you must redeploy the bundle and rerun the agent entry after adding new MCP servers or updating permissions in databricks.yml. This post-change deployment ensures the runtime environment reflects the new tool wiring and access grants.

Can I connect custom MCP apps to my LangGraph agent using this approach?

Yes, you can connect custom mcp-* app-based MCP servers to your LangGraph agent. You load their tool definitions into the agent and provision the corresponding permissions in databricks.yml to allow runtime execution.