Quickstart

Prerequisites

• A running ctx| deployment with a public or local AUTH_BASE_URL
• A GitHub repository
• Cursor, Claude Code, or any MCP-compatible agent

1. Start the stack

For a production-style local run, use the Compose deploy profile from the repo root:

docker compose --profile deploy up -d

Open the app at your configured AUTH_BASE_URL. For the default local Compose ports, that is http://localhost:3000; the backend proxies the UI from there.

2. Complete onboarding and connect a repository

On first sign-in, complete onboarding to create or select your organization. Then navigate to Repositories and add your repo URL. ctx| will clone and index your codebase - this typically takes 1-5 minutes depending on size.

For GitHub-hosted code, use the GitHub App connector when available. Self-hosted deployments can store GitHub App credentials per connection; see the GitHub App section on Configuration.

3. Configure your MCP client

Add the following to your MCP client configuration (e.g. .cursor/mcp.json):

{
  "mcpServers": {
    "ctxpipe": {
      "type": "streamable-http",
      "url": "https://ctx.example.com/mcp?orgSlug=your-org-slug"
    }
  }
}

Replace your-org-slug with the slug shown in Settings → Organization. Replace https://ctx.example.com with your deployment's AUTH_BASE_URL. Your MCP client should complete OAuth authorization against that deployment the first time it connects.

4. Start your agent

Your agent now has access to the knowledge graph. It will automatically query ctx| via the ctx_advisor tool during planning and execution to understand your architecture and follow your standards.

For best results, ensure your agent calls ctx_advisor early in any new task, before writing code, and again whenever a major decision changes.

Related setup

• Architecture - services, state, and request flow.
• Configuration - required environment variables and connector secrets.
• Authentication - sign-in, API keys, and OAuth.
• Model configuration - LLM tiers and embeddings.
• Graph databases - FalkorDB, Neo4j, Memgraph, and Neptune.
• Operations - health checks, logs, backups, and recovery.
• Confluence & Atlassian - self-hosted Confluence connector setup.