ContextKit

$ npm install -g @runcontext/clicopy

How It Works

You tell your AI agent to build it. The agent runs CLI commands, asks you about your data, and iterates until Gold. You provide the business knowledge. The agent does the rest.

Without ContextKit

— “What’s total revenue?”
SELECT AVG(revenue)     ← wrong aggregation
FROM transactions       ← no status filter
← includes cancelled

With ContextKit

— “What’s total revenue?”
SELECT SUM(revenue)     ← correct (semantic role)
FROM transactions
WHERE status=‘completed’ ← required (guardrail)

Three Tiers to AI-Ready

Bronze — Discoverable

Descriptions, ownership, security, grain. Achieved automatically via context introspect.

Silver — Trusted

Trust status, glossary, lineage, sample values. Achieved via context enrich --target silver.

Gold — AI-Ready

Semantic roles, aggregation rules, guardrails, golden queries, business rules. Curated by you and your agent.

40 Lint Rules

Schema validation, governance, security, data accuracy — all tier-aware with auto-fix.

Everything in One Install

16 CLI Commands

setup, new, introspect, enrich, lint, fix, build, tier, blueprint, explain, verify, serve, site, dev, init, rules, validate-osi

MCP Server

Serve your semantic layer to Claude Code, Cursor, Windsurf, and any MCP-compatible agent via stdio or HTTP.

Visual Studio Editor

context dev --studio — edit metadata in the browser, see tier updates live, save back to YAML.

9 Database Adapters

PostgreSQL, DuckDB, MySQL, SQL Server, Snowflake, BigQuery, ClickHouse, Databricks, SQLite.

Quick Start

Tell your AI agent: “Install @runcontext/cli and build a semantic layer for my database.”

Or do it yourself:

npm install -g @runcontext/cli
context setup           # connect → introspect → scaffold → enrich → generate agent instructions
context tier            # check your Bronze/Silver/Gold score
context serve --stdio   # serve to AI agents via MCP

Introduction How agents use ContextKit

Quick Start Up and running in 5 minutes

OSI Specification The open standard behind it

MCP Server Serve context to AI agents