Database Support

ContextKit connects to your database to introspect schemas, validate metadata against live data, and verify golden queries. It supports a wide range of databases through optional peer dependencies — install only the driver you need.

Supported databases

PostgreSQL

Also works with Neon, Supabase, Redshift, CockroachDB, AlloyDB, and any Postgres wire-compatible database.

DuckDB

Local .duckdb files, in-memory databases, Parquet/CSV/JSON via DuckDB views. Also works with MotherDuck.

MySQL / MariaDB

MySQL 5.7+, 8.x, MariaDB 10.x+. Works with PlanetScale, TiDB, and other MySQL-compatible services.

SQL Server

Microsoft SQL Server 2016+. Also works with Azure SQL Database.

Snowflake

Full support for Snowflake accounts with warehouse/database/schema selection.

BigQuery

Google BigQuery with service account authentication and dataset selection.

ClickHouse

ClickHouse HTTP interface. Works with ClickHouse Cloud.

Databricks

Databricks SQL warehouses via server hostname and HTTP path.

SQLite

Local .sqlite / .db files for lightweight or embedded databases.

Quick setup

The fastest way to connect is the interactive wizard:

context setup

It will auto-detect databases from your environment, MCP configs (Claude Code, Cursor, VS Code, Windsurf), and contextkit.config.yaml.

Configuration

Connection string databases

PostgreSQL, MySQL, and SQL Server use standard connection strings:

PostgreSQL

contextkit.config.yaml

data_sources:
  warehouse:
    adapter: postgres
    connection: postgresql://user:pass@host:5432/dbname

Also works for Neon, Supabase, Redshift, CockroachDB — they all speak the PostgreSQL wire protocol.

MySQL

contextkit.config.yaml

data_sources:
  warehouse:
    adapter: mysql
    connection: mysql://user:pass@host:3306/dbname

Install the driver: npm install mysql2

SQL Server

contextkit.config.yaml

data_sources:
  warehouse:
    adapter: mssql
    connection: mssql://user:pass@host:1433/dbname

Install the driver: npm install mssql

File-based databases

DuckDB

contextkit.config.yaml

data_sources:
  analytics:
    adapter: duckdb
    path: ./data/warehouse.duckdb

DuckDB is included by default — no extra install needed.

SQLite

contextkit.config.yaml

data_sources:
  local:
    adapter: sqlite
    path: ./data/app.sqlite

Install the driver: npm install better-sqlite3

Cloud warehouses

Snowflake

contextkit.config.yaml

data_sources:
  snowflake:
    adapter: snowflake
    account: xy12345.us-east-1
    username: analyst
    password: ${SNOWFLAKE_PASSWORD}
    warehouse: COMPUTE_WH
    database: ANALYTICS
    schema: PUBLIC

Install the driver: npm install snowflake-sdk

BigQuery

contextkit.config.yaml

data_sources:
  bigquery:
    adapter: bigquery
    project: my-gcp-project
    dataset: analytics
    keyFilename: ./service-account.json

Install the driver: npm install @google-cloud/bigquery

ClickHouse

contextkit.config.yaml

data_sources:
  clickhouse:
    adapter: clickhouse
    host: localhost
    port: 8123
    database: analytics
    username: default
    password: ${CLICKHOUSE_PASSWORD}

Install the driver: npm install @clickhouse/client

Databricks

contextkit.config.yaml

data_sources:
  databricks:
    adapter: databricks
    serverHostname: dbc-abc123.cloud.databricks.com
    httpPath: /sql/1.0/warehouses/abcdef
    token: ${DATABRICKS_TOKEN}

Install the driver: npm install @databricks/sql

Caution

Never commit credentials to version control. Use ${ENV_VAR} syntax in your config — ContextKit expands environment variables at runtime.

Driver installation

ContextKit uses optional peer dependencies — install only the driver for your database:

Database	Driver Package	Install Command
DuckDB	duckdb	Included by default
PostgreSQL	pg	npm install pg
MySQL / MariaDB	mysql2	npm install mysql2
SQL Server	mssql	npm install mssql
Snowflake	snowflake-sdk	npm install snowflake-sdk
BigQuery	@google-cloud/bigquery	npm install @google-cloud/bigquery
ClickHouse	@clickhouse/client	npm install @clickhouse/client
Databricks	@databricks/sql	npm install @databricks/sql
SQLite	better-sqlite3	npm install better-sqlite3

If a driver is missing, ContextKit will tell you exactly which package to install.

MCP config auto-detection

When you run context setup, ContextKit scans your IDE MCP configs to find databases you’ve already connected:

IDE	Config Location
Claude Code	~/.claude.json, <project>/.mcp.json
Cursor	~/.cursor/mcp.json, <project>/.cursor/mcp.json
VS Code / Copilot	<project>/.vscode/mcp.json
Windsurf	~/.codeium/windsurf/mcp_config.json
Claude Desktop	~/Library/Application Support/Claude/claude_desktop_config.json

If a database MCP server is found, context setup offers it as the first option — no connection string needed.

Introspect via CLI

For non-interactive use, pass a connection string directly:

# PostgreSQL (also Neon, Supabase, Redshift)
context introspect --db postgres://user:pass@host:5432/db

# DuckDB
context introspect --db ./warehouse.duckdb

# MySQL
context introspect --db mysql://user:pass@host:3306/db

# SQL Server
context introspect --db mssql://user:pass@host:1433/db

# Snowflake (via config)
context introspect --source snowflake

# BigQuery (via config)
context introspect --source bigquery

Cloud warehouses (Snowflake, BigQuery, ClickHouse, Databricks) require multiple credentials and should be configured in contextkit.config.yaml rather than passed as a single URL.

Multiple data sources

Configure multiple databases if your models span different systems:

contextkit.config.yaml

data_sources:
  events:
    adapter: clickhouse
    host: click.example.com
    database: events
  users:
    adapter: postgres
    connection: ${USERS_DB_URL}
  analytics:
    adapter: duckdb
    path: ./data/analytics.duckdb

Each model’s source field determines which data source is used for validation and introspection.