Alpha
Engineering Intelligence
Know the true state of your engineering. Point Alpha at a codebase — a GitHub repo, a website, or a ZIP — and get a grounded report: architecture, security findings, dependency risk, and code quality, backed by a knowledge graph and an AI you can question.
Overview
Alpha turns a repository or live website into a relational knowledge graph, then layers intelligence on top: health scores, predictive risk, team velocity, a grounded AI you can question, and autonomous agents that draft fixes for approval. Analysis is the data source; intelligence is the product.
Features
Analysis Engine (AST-level understanding across 40+ languages), Architecture Maps (living dependency diagrams), Security Audit (context-aware CVE ranking against your actual reachable code), API Generation (OpenAPI specs from your routing layer), AI Assistant (chat grounded in your own repository, zero hallucination by construction), and deep website scans — headers, cookies, forms, TLS, and IDE-grade source reconstruction from source maps.
Pricing
Free to start with the full analysis engine, health scores, security findings, and AI chat. Paid tiers unlock unlimited analyses, private repositories, teams, and autonomous agents — see /pricing for current plans.
Documentation
Full docs at /docs, including the API reference at /api-docs and MCP setup at /docs/mcp.
Roadmap
Next up: interactive architecture diagrams, scheduled recurring analyses, compare mode between two analyses, and private-repository support.
Integrations
GitHub, GitLab, and live websites as analysis sources; the MCP Hub for AI-assistant access; API keys for the CLI and CI; webhook delivery for analysis events.
FAQ
What is Norveth?
Norveth is an AI-native engineering intelligence platform. It analyses your repositories, websites, or ZIP uploads into a living knowledge graph, then layers intelligence on top: engineering health scores, predictive risk analysis, team velocity dashboards, a grounded AI CTO you can question, and autonomous agents that draft fixes for your approval.
What is an engineering intelligence platform?
Where a code scanner produces a one-off report, an engineering intelligence platform continuously measures the state of your engineering — architecture, security, dependencies, code quality, team velocity, and contributor risk — across every repository, and turns it into scores, predictions, and actions leaders can use.
What can Norveth analyse?
Public and private GitHub repositories (via the GitHub App), live websites by URL, and ZIP uploads. Analysis covers architecture, security findings including real CVEs, dependency risk, code quality, performance, and auto-generated documentation — grounded in the real code, not a generic AI guess.
What do Norveth's autonomous agents do?
Agents act on what the analysis finds: AutoFix drafts dependency upgrades and vulnerability remediation plans, and AutoDocument generates missing documentation for complex code paths. Every agent plan requires human approval before it executes, and every action is recorded in the audit log.
Is Norveth free to use?
Yes — the Free plan includes the full analysis engine with monthly limits and no credit card required. Paid plans add volume and advanced capabilities: unlimited analyses, private repositories, team workspaces, predictive insights, and autonomous agents.
Can I use Norveth inside Claude, ChatGPT, or Gemini?
Yes. Norveth ships a Model Context Protocol (MCP) server: connect the hosted endpoint to Claude or ChatGPT on the web, or run the @norveth/mcp bridge locally for Claude Desktop, Claude Code, Cursor, Windsurf, and any other MCP client. Authenticate once with your Norveth API key and your assistant can analyze repositories, read reports, and ask grounded questions about your codebase.
How is Norveth different from asking ChatGPT or Claude about my code?
A generic AI chat has no access to your actual codebase and can hallucinate. Norveth builds an engineering knowledge graph from your files first, then its AI CTO answers questions grounded in that graph — every answer traces back to actual files, dependencies, and metrics, not guesses.