Skip to content
RunContext Docs

context setup

The setup command is the fastest way to get started with RunContext. It opens a 6-step wizard in your browser that creates a semantic plane from scratch.

Usage

Section titled “Usage”
Terminal window
context setup
context setup --port 4040      # custom port (default: 4040)
context setup --no-browser     # don't auto-open the browser

What it does

Section titled “What it does”

The setup wizard opens at http://localhost:4040/setup and walks through six steps:

Step 1: Connect

Section titled “Step 1: Connect”

Connect to your database. Data sources are auto-detected from environment variables (DATABASE_URL, POSTGRES_URL, etc.), local database files (.duckdb, .sqlite), and MCP configs from your IDE.

Step 2: Define

Section titled “Step 2: Define”

Name your semantic plane and provide a description. Set the owner name, team, and email. Choose a sensitivity level (public, internal, confidential, or restricted). The name becomes the directory under context/products/.

Step 3: Scaffold

Section titled “Step 3: Scaffold”

Introspect the database schema and generate Bronze-tier OSI YAML files. This reads tables, columns, types, primary keys, and relationships.

Step 4: Checkpoint

Section titled “Step 4: Checkpoint”

Review the scaffolded metadata before enrichment. Edit any field before proceeding.

Step 5: Curate

Section titled “Step 5: Curate”

Agent-driven curation enriches your metadata from Bronze through Silver and Gold:

  1. Enrich to Silver — add glossary, lineage, sample values, trust status
  2. Enrich to Gold — suggest semantic roles, aggregation rules, guardrails
  3. Verify — lint and validate the result
  4. Auto-fix — apply automatic fixes for common issues
  5. Agent Instructions — generate AGENT_INSTRUCTIONS.md for AI curation

Step 6: Serve

Section titled “Step 6: Serve”

Start the MCP server and configure your AI tools to connect. The wizard generates the appropriate config snippets for Claude Code, Cursor, VS Code, and other MCP clients.

Output

Section titled “Output”

The wizard saves a Context Brief at context/products/{name}/context-brief.yaml and scaffolds all associated metadata files for the semantic plane.

context-brief.yaml
product_name: sales-analytics
description: "Revenue and order metrics for the analytics team"
owner:
  name: Alice Chen
  team: Analytics
  email: alice@company.com
sensitivity: internal
created_at: "2026-03-08T..."

When to use setup vs. individual commands

Section titled “When to use setup vs. individual commands”
ScenarioRecommendation
First time with RunContextcontext setup (browser wizard)
Adding a new database sourcecontext introspect --db <url>
Upgrading existing context to Goldcontext dev --studio (visual editor)
CI/CD pipelinecontext lint && context build
Let an AI agent build itTell the agent: “Install @runcontext/cli and build a semantic plane”

Next steps

Section titled “Next steps”