v1.0.0 · MIT

Super Codebase
Context.

One command. Full AI coding stack. Auto-detects your project. Configures semantic search, architectural boundary enforcement, and a verify loop. Any language. Any team size.

bash

$ curl -fsSL https://raw.githubusercontent.com/MasoomChoudhury/super-codebase-context/main/install.sh | bash

✓ Detected: TypeScript · Next.js · pnpm

✓ .context/index.md generated

✓ Agent instructions generated (Claude, Cursor, Windsurf)

✓ cocoindex-code installed (local · no API key)

✓ Project indexed → mcp registered in IDE

✓ polyglot verify-tools registered (tsc, mypy, clippy)

✓ pre-commit hook installed

── Done in 89 seconds ──

copy install command

View on GitHub Quick Start →

< 2 min

Full stack install time

Architecture layers

API keys required

polyglot

Support (JS, PY, GO, RS)

local

Embeddings & logic

Architecture

Six layers, zero config

A multi-agent, production-grade stack tailored to your project structure and language.

0Layer

Codebase Context Specification

.context/index.md .contextignore per-module stubs

The highest-ROI layer. Based on the CCS spec — works like .editorconfig for AI tools. Every agent reads this first to understand your module boundaries and conventions without you re-explaining every session.

highest ROI

1Layer

Multi-Agent Instructions

CLAUDE.md .cursorrules AGENTS.md copilot.md

Auto-generates cross-IDE persistent instructions. Whether you use Claude, Cursor, Windsurf, or Copilot, your AI gets the exact same system prompt with correct build commands and module rules already filled in.

cross-IDE

2Layer

Semantic Index (Local MCP)

cocoindex-code ccc mcp search_codebase()

Free, fully local semantic indexing. Uses on-device embeddings to answer queries like "where does rate limiting happen?" Handles millions of lines of code without cloud dependency.

fully local

3Layer

Polyglot Verify Loop (MCP)

verify_syntax() run_tests() self-correct

A custom local MCP server exposing your project's toolchain. AI calls these before presenting output — self-correcting via tsc, mypy, golangci-lint, or cargo based on your detected language.

verify → correct

4Layer

Precision Intelligence (Optional)

cased-kit tree-sitter symbols

Interactive opt-in for deterministic codebase mapping. Extracts surgical AST-aware symbols and usages instead of probabilistic search. Massive token savings by giving agents maps instead of raw files.

opt-in

5Layer

Safety Net (Pre-commit)

dependency-cruiser import-linter git hook

Hard boundary enforcement. Defines rules like "UI must not import from API." Pre-commit hook runs auto-detected linters and boundary checks on every commit. Fails if architecture is violated.

hard boundaries

MCP Tools

What your AI can call

These tools are registered in your IDE automatically. Your AI calls them on its own to explore and verify code.

Tool

What it runs

When called

search_codebase(query)

Semantic search across entire codebase via cocoindex-code (local, free)

Before reading files — finds code by meaning

verify_syntax()

Runs tsc / mypy / golangci / clippy + linter, returns errors with line numbers

After every code generation step

run_tests(pattern?)

Runs test suite (npm, pytest, cargo, go test) scoped to affected files

After touching test-covered code

check_deps()

Runs dependency-cruiser or import-linter for boundary violations

Before refactors touching imports

Output

Unified Context Infrastructure

Every teammate gets the identical AI setup the moment they git pull.

📋

.context/index.md

Architecture overview and module boundaries. Edit this first.

🤖

CLAUDE.md / .cursorrules

Cross-IDE instructions with detected build commands and conventions.

🏗️

.dependency-cruiser.cjs

Architectural boundary rules. Pre-configured for UI→API constraints.

🔒

.git/hooks/pre-commit

Polyglot hook for tsc, mypy, clippy, and boundary checks.

⚡

.scc/mcp-verify/

Local Node.js MCP server exposing your project's toolchain as tools.

🚫

.contextignore

Excludes node_modules, build, and dist from AI context.

Compatibility

Agent-Agnostic Context

The installer automatically configures your primary tool's mcp.json or settings.

Claude Code

✓ CLAUDE.md · CLI Auto

Cursor

✓ .cursorrules · mcp.json

Windsurf

✓ AGENTS.md · mcp_config

Gemini CLI

✓ Context · Manual MCP

GitHub Copilot

✓ instructions · Hook

Quick Start

Production setup in seconds

Run the installer

Works on macOS, Linux, and WSL. Automatically detects your tech stack.

$ curl -fsSL https://raw.githubusercontent.com/MasoomChoudhury/super-codebase-context/main/install.sh | bash

Define Boundaries in .context/

The installer scaffolds index files for major modules. Fill in the "What NOT to import" section to stop AI hallucinations.

Select your IDE

The script will ask if you use Claude, Cursor, or Windsurf and write the local MCP config for you.

Commit and Share

Commit the generated files. Teammates get the full stack instantly on their next pull.

$ git add .context/ CLAUDE.md .cursorrules .dependency-cruiser.cjs .contextignore
$ git commit -m "chore: add AI coding stack"

Super CodebaseContext.