GitHub - nxtg-ai/repoatlas: Portfolio intelligence for AI engineering teams. Scan repos, score health, find cross-project patterns.

5 min read Original article ↗

nxtg-atlas

PyPI Python License: MIT Tests

Portfolio intelligence for AI engineering teams.

Scan your repos. Score health. Find cross-project patterns nobody else sees.

Atlas terminal demo


The Problem

You manage 5, 10, 20+ repos. Every tool on the market — Cursor, Copilot, Claude Code — works inside a single repo. Nobody sees the full picture:

  • Which repos have zero tests?
  • Are your React versions consistent across projects?
  • Where are you duplicating FastAPI patterns instead of sharing them?
  • Which repos haven't been touched in weeks?

You don't know until something breaks.

The Solution

atlas init --name "My Portfolio"
atlas add ~/projects/api
atlas add ~/projects/frontend
atlas add ~/projects/ml-pipeline
atlas scan
atlas status

30 seconds. Full portfolio health dashboard. Real data from your repos.

What You Get

╭───────────────────────  ATLAS Portfolio Dashboard  ─────────────────────────╮
│                                                                             │
│  Portfolio: NXTG.AI  |  8 Projects  |  609 Test Files  |  1,251,303 LOC    │
│  Health: B (82%)  |  Scanned: 2026-03-04T01:15                             │
│                                                                             │
╰─────────────────────────────────────────────────────────────────────────────╯
┌────┬────────────────────┬──────────┬────────┬──────────┬─────────────────────┐
│    │ Project            │ Health   │  Tests │      LOC │ Tech Stack          │
├────┼────────────────────┼──────────┼────────┼──────────┼─────────────────────┤
│ ●  │ Dx3                │  A 94%   │    179 │  164,806 │ Python · TypeScript │
│ ●  │ content-engine     │  B+ 88%  │     23 │   12,552 │ Python              │
│ ●  │ Podcast-Pipeline   │  B+ 88%  │     49 │   33,680 │ Python · FastAPI    │
│ ●  │ nxtg.ai            │  B+ 87%  │     36 │   49,938 │ TypeScript · Next.js│
│ ●  │ Faultline          │  B+ 86%  │     29 │   15,527 │ TypeScript · React  │
│ ●  │ voice-jib-jab      │  B+ 86%  │     58 │   45,311 │ TypeScript · Python │
│ ●  │ SynApps            │  C 67%   │    114 │  793,406 │ Python · TypeScript │
│ ●  │ NXTG-Forge         │  D 59%   │    121 │  136,083 │ TypeScript · Rust   │
└────┴────────────────────┴──────────┴────────┴──────────┴─────────────────────┘

This is real output from scanning 8 production repos. Not a mockup.

Features

  • Health scoring — 4 dimensions (tests, git hygiene, documentation, structure) → A-F grade
  • Tech stack detection — Python, TypeScript, Rust, Go, Java + frameworks (FastAPI, React, Next.js, Django, etc.)
  • Git health — branch, commit count, dirty state, remote status
  • Database detection — PostgreSQL, SQLite, Redis, MongoDB, pgvector
  • Cross-project intelligence — shared deps, version mismatches, health gaps
  • Batch addatlas batch-add ~/projects adds every repo at once
  • Export — markdown and JSON reports for stakeholders
  • Unlimited projects — no limits, no tiers, no gates

Everything is free. Everything is open source. No catches.

What It Detects

Category Examples
Languages Python, TypeScript, JavaScript, Rust, Go, Java, Ruby, C++, Swift, Kotlin
Frameworks FastAPI, Django, Flask, React, Next.js, Vue, Express, Tailwind, Vite
Test Frameworks pytest, Vitest, Jest, Playwright
Databases PostgreSQL, SQLite, Redis, MongoDB, pgvector
CI/CD GitHub Actions, GitLab CI, Jenkins, CircleCI
Infrastructure Docker, Kubernetes, Helm, Terraform, Pulumi, AWS CDK
Cloud AWS, GCP, Azure
Serverless Vercel, Netlify, Cloudflare Workers, Fly.io, Render
Security Dependabot, Renovate, Snyk, CodeQL, Bandit, Gitleaks, Trivy, SECURITY.md
AI/ML Anthropic SDK, OpenAI, LangChain, LlamaIndex, Transformers, PyTorch, TensorFlow, scikit-learn, Vercel AI SDK, MLflow, W&B, DVC, ChromaDB, Pinecone, Jupyter
Code Quality Ruff, Flake8, Pylint, ESLint, Biome, Black, Prettier, isort, mypy, Pyright, TypeScript, golangci-lint, Clippy
Documentation README, CLAUDE.md, CHANGELOG, docs/

Health Score Breakdown

Each project scores 0-100% across 4 dimensions:

  Health Breakdown:
    Tests          █████████████░░  90%
    Git Hygiene    ██████████████░  99%
    Documentation  ███████████████  100%
    Structure      █████████████░░  90%
Dimension What It Measures
Tests (35%) Test file count relative to source files
Git Hygiene (20%) Commits, remote, clean working tree
Documentation (20%) README, CLAUDE.md, NEXUS, CHANGELOG, docs/
Structure (25%) CI/CD, .gitignore, package config, source organization, linting

Atlas finds patterns across your repos that no single-repo tool can see:

╭───────────────── Cross-Project Intelligence ──────────────────╮
│                                                                │
│  Shared Dependencies                                           │
│    ℹ  fastapi used across 4 projects                           │
│    ℹ  react used across 3 projects                             │
│                                                                │
│  Version Mismatches                                            │
│    ⚠  react: ^18.2.0 (app-a), ^19.2.1 (app-b)                 │
│    ⚠  fastapi: >=0.100.0 (api-v1), >=0.115.9 (api-v2)         │
│                                                                │
│  Health Gaps                                                   │
│    ✖  3 projects have zero tests                               │
│    ⚠  1 project has 50+ uncommitted changes                    │
│                                                                │
╰────────────────────────────────────────────────────────────────╯

Commands

Command Description
atlas init Create a portfolio
atlas add <path> Add a project
atlas scan Re-scan all projects
atlas status Portfolio dashboard
atlas inspect <name> Deep-dive on one project
atlas compare <a> <b> Side-by-side project comparison
atlas remove <name> Remove a project
atlas connections Cross-project patterns
atlas doctor Diagnose issues and suggest fixes
atlas trends Show health trends across scans
atlas ci Health gate for CI pipelines (JSON output, exit codes)
atlas batch-add <dir> Add all repos in a directory
atlas export Markdown/JSON report
atlas config View or update configuration
atlas support How to support this project
atlas reset Delete all data

Requirements

  • Python 3.11+
  • git (for repo detection)

How It Works

Atlas scans your project directories — no network calls, no cloud, no telemetry. Everything runs locally.

  1. File walking — counts source and test files, skipping node_modules, .venv, target, etc.
  2. Config parsing — reads pyproject.toml, package.json, Cargo.toml for deps and frameworks
  3. Git inspection — runs git log, git status, git branch for repo health
  4. Health scoring — weighted formula across 4 dimensions → A-F grade
  5. Cross-project analysis — compares deps, versions, and patterns across all repos

State lives in ~/.atlas/portfolio.json. Portable, inspectable, no database.

FAQ

Is this open source? Yes. MIT license. Every feature is free. No tiers, no gates, no "upgrade to unlock."

Does it phone home? No. Zero network calls. No telemetry. No analytics. Your code stays on your machine.

How fast is it? 8 projects with 1.25M lines of code scanned in 31 seconds.

Does it work with monorepos? Each atlas add path is treated as one project. For monorepos, add the root.

Can I use it in CI? Yes. atlas ci outputs JSON and exits non-zero on health violations:

# GitHub Actions example
- run: atlas ci --min-health 70 --min-project-health 50

Support This Project

Atlas is free and always will be. If it saves you time, consider supporting development:

Support Atlas →

Supporters get their name in SUPPORTERS.md, priority GitHub issues, and early access to new features.

License

MIT


Built by NXTG.AI — the team behind NXTG-Forge.