data-context-mcp

Standalone MCP (Model Context Protocol) server for data context and catalog operations. This is the free-tier OSS entry point for Data Workers -- giving AI assistants deep understanding of your data assets, lineage, documentation, and impact analysis.

Quick Start

npx data-context-mcp

No configuration needed. The server starts with in-memory backends seeded with sample data so you can explore immediately.

Claude Desktop Integration

Add this to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "data-context": {
      "command": "npx",
      "args": ["data-context-mcp"]
    }
  }
}

Then ask Claude things like:

• "What tables do we have related to orders?"
• "Trace the lineage of the revenue_daily table"
• "What would break if I changed the customers table?"
• "Explain the orders table"

Tools

Read-Only Tools (Free -- Community Tier)

All 20 read-only tools are available without any license:

| Tool | Description | |------|-------------| | search_datasets | Natural language search over the data catalog | | search_across_platforms | Federated cross-platform search with hybrid retrieval | | get_lineage | Column-level lineage traversal | | trace_cross_platform_lineage | Unified cross-platform lineage tracing | | resolve_metric | Metric name disambiguation via semantic layer | | list_semantic_definitions | Browse semantic layer definitions | | get_documentation | Retrieve auto-generated asset documentation | | check_freshness | Data freshness and SLA compliance checks | | get_context | Complete context for a data asset in a single call | | explain_table | Hero tool -- full table explanation with schema, lineage, quality | | assess_impact | Downstream impact assessment for asset changes | | blast_radius_analysis | Column-level, cross-platform blast radius with PR diff support | | detect_dead_assets | Find orphaned tables, columns, and DAG nodes | | correlate_metadata | Cross-platform metadata enrichment and correlation | | get_table_schema_for_sql | Schema info optimized for SQL generation | | query_rules | Search and browse business rules | | get_authoritative_source | Look up authoritative source for a domain/concept | | analyze_query_history | Analyze query patterns and usage history | | identify_golden_path | Identify recommended query paths for data access | | check_staleness | Check if context/documentation is stale |

Write Tools (Pro Tier)

These 11 tools require DW_LICENSE_TIER=pro (or higher):

| Tool | Description | |------|-------------| | generate_documentation | Generate and persist documentation with provenance | | update_lineage | Additive lineage edge management with soft-delete | | auto_tag_dataset | Auto-classify datasets with tags (with rollback) | | flag_documentation_gap | Flag missing or stale documentation | | define_business_rule | Create business rules for data assets | | import_tribal_knowledge | Batch import tribal knowledge as business rules | | update_business_rule | Update existing business rules | | mark_authoritative | Designate an asset as authoritative for a domain | | revoke_authority | Revoke authority designation | | correct_response | Submit corrections to context/documentation | | flag_stale_context | Flag context as stale with reason |

Admin Tools (Enterprise Tier)

These 2 tools require DW_LICENSE_TIER=enterprise:

| Tool | Description | |------|-------------| | ingest_unstructured_context | Ingest unstructured context from Slack, Confluence, etc. | | run_data_steward | Enterprise data steward workflow automation |

Environment Variables

License Tier

| Variable | Values | Default | |----------|--------|---------| | DW_LICENSE_TIER | community, pro, enterprise | community |

Data Platform Connectors

Set these to connect to your real data infrastructure instead of in-memory backends:

| Variable | Description | |----------|-------------| | SNOWFLAKE_ACCOUNT | Snowflake account identifier | | SNOWFLAKE_USERNAME | Snowflake username | | SNOWFLAKE_PASSWORD | Snowflake password | | SNOWFLAKE_WAREHOUSE | Snowflake warehouse name | | SNOWFLAKE_DATABASE | Snowflake database name | | GCP_PROJECT | Google Cloud project ID (for BigQuery) | | GOOGLE_APPLICATION_CREDENTIALS | Path to GCP service account key | | DBT_PROJECT_DIR | Path to dbt project directory | | DBT_PROFILES_DIR | Path to dbt profiles directory | | DATABRICKS_HOST | Databricks workspace URL | | DATABRICKS_TOKEN | Databricks personal access token | | REDSHIFT_HOST | Amazon Redshift cluster endpoint | | REDSHIFT_DATABASE | Redshift database name | | REDSHIFT_USER | Redshift username | | REDSHIFT_PASSWORD | Redshift password |

How It Works

This package bundles the dw-context-catalog agent from the Data Workers platform as a standalone MCP server. It communicates via JSON-RPC 2.0 over stdio (stdin/stdout), which is the standard MCP transport for local tools.

• Community tier (default): All read-only tools work out of the box. Write tools return an upgrade prompt.
• Pro tier: Unlocks write tools for documentation generation, lineage management, tagging, and business rules.
• Enterprise tier: Adds admin tools for unstructured context ingestion and data steward workflows.

License

Apache-2.0