What it does
X-Ray extracts the architecture of a repository without an LLM, so the code stays on the local machine. It writes an ARCHITECTURE.md map, an AGENT_CONTEXT.md file for rules files such as CLAUDE.md, an architecture.json model and a diagrams folder with Mermaid sources and an HTML viewer. For Python it reads Flask and FastAPI routes with auth-guard detection, SQLAlchemy and Django models, Celery tasks and the import graph. For JS and TS it covers Next.js, Express routes and Prisma models. It also flags hardcoded secrets, raw SQL that bypasses the ORM, auth tokens in localStorage or AsyncStorage, and routes without auth guards. The production-readiness score covers eight dimensions and lists the top fixes ranked by points lost.
It suits reviewing AI-generated code before it ships, auditing an inherited codebase, checking how far a repository has drifted over time, and checking the impact of a change before an agent edits a file. The score is deterministic, so it can be compared across commits and run in CI.
To use it, install the Claude Desktop extension or the Official MCP Registry listing, and provide the repository to analyse. The MCP tools are blast_radius, arch_summary, boundary_findings, production_readiness, generate_diagrams and xray_scan, which re-scans after structural changes. The tool uses only the standard library and makes no network calls.
Limitations and requirements
Extraction covers only the stacks listed: Python, JS and TS. Other languages such as Go, Java, Rails and .NET are not yet supported. Auth status can be reported as unknown (?) when auth is attached at router level, because per-function analysis cannot see it. The production-readiness score is a static sign-off aid and not a security audit. Using the Claude Desktop extension requires choosing the repository to analyse at install time.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
io.github.Anioko/archiet-xray
Deterministic repo architecture maps: auth coverage, blast radius, readiness score, diagrams
v0.2.3As shown on registry.modelcontextprotocol.io when we last read it