DataBridge Plugin for Claude Code

Data reconciliation, hierarchy management, and financial modeling — packaged as a Claude Code plugin.

Upload a Chart of Accounts. Get a production-ready financial hierarchy and dbt models. Zero config.

Installation

From local directory (development)

claude --plugin-dir ./databridge-plugin

From GitHub (once published)

claude plugin install datanexum/databridge-plugin

Prerequisites

• Python 3.10+
• Core dependencies: pip install pandas pydantic fastmcp
• Optional: pip install 'databridge-core[all]' for full feature set

Run /databridge:setup after installation to verify everything is configured.

What's Included

Skills (auto-invoked by Claude based on context)

Skill	Triggers on	What it does
onboard	New CSV/Excel upload	Load → profile → recommend → optional fuzzy match
reconcile	"compare", "reconcile", "diff"	Hash compare → orphans → conflicts → fuzzy match → summary
triage	Directory of files, "scan"	Classify archetypes, detect errors, recommend processing order
profile	"profile", "quality", "validate"	Statistical profiling → expectations → validation
hierarchy	"hierarchy", "COA", "chart of accounts"	Multi-level hierarchy creation with formulas
model	"model", "dbt", "data mart"	Model discovery → dbt generation → mart pipeline

Commands (user-invoked)

Command	Description
/databridge:setup	Check dependencies, test connections, list tools
/databridge:status	Server health, tool count, license tier, audit log

Agents (specialized sub-agents for complex workflows)

Agent	Description
data-quality	Profiles sources, detects drift, validates expectations, produces quality scorecards
reconciliation	Full reconciliation lifecycle: load → compare → analyze → fuzzy match → merge → report
financial-modeler	Hierarchy creation → source mapping → formulas → mart generation → dbt → deployment

Hooks

• PostToolUse: After CSV import, suggests running /databridge:profile for quality checks

MCP Server

The plugin configures a DataBridge MCP server with 316 CE tools (up to 381 with Pro/Enterprise) in full mode. The server starts automatically when Claude Code loads the plugin.

Tools are organized across 46 categories including:

• Data profiling and quality
• Reconciliation (hash, fuzzy, diff)
• Hierarchy management
• Data modeling and dbt
• Snowflake and database connectors
• BLCE engine (84 tools)
• Mart factory
• Catalog and lineage

Alternative: Docker-based MCP Server

If you prefer running the MCP server in Docker instead of locally:

# Start the MCP server container
docker compose --profile mcp up databridge-mcp -d

# Or build and run directly
docker build -f Dockerfile.mcp -t databridge-mcp .
docker run -p 786:786 databridge-mcp

Then configure your LLM client to connect via SSE:

{
  "mcpServers": {
    "DataBridge AI": {
      "url": "http://localhost:786/sse"
    }
  }
}

Configuration

Environment Variables

Variable	Default	Description
DATABRIDGE_TOOL_MODE	full	Tool loading mode: slim, dynamic, full
DATABRIDGE_DEMO_MODE	false	Enable demo mode with sample data
DATABRIDGE_TELEMETRY	true	Local telemetry collection

Custom MCP Server Path

If your DataBridge installation is not in the parent directory of the plugin, update .mcp.json:

{
  "mcpServers": {
    "databridge": {
      "command": "python",
      "args": ["run_server.py", "--full"],
      "cwd": "/path/to/your/DATABRIDGE_AI"
    }
  }
}

Quick Start

1. Install the plugin
2. Run /databridge:setup to verify dependencies
3. Drop a CSV file and say "profile this data"
4. Say "build a hierarchy from this chart of accounts"
5. Say "generate a dbt project for this model"

License

MIT

Links

• DataBridge Core (PyPI)
• GitHub