ArchiLens reads your Git repository and generates layered C4-style architecture diagrams — automatically, using AST analysis and AI enrichment. No manual diagrams. No stale docs.
From a single CLI command to a full interactive viewer — ArchiLens covers the whole workflow.
Tree-sitter parses Python, JavaScript, TypeScript, Java, and Go at the AST level, with a regex fallback for other languages. Classes, functions, imports, and inheritance are extracted with line-level precision. Imports are then resolved to files per language — Python packages and src/ layouts, TypeScript path aliases and workspace packages, Java packages, Go modules — so dependency edges reflect real imports.
Optionally, Claude writes a one-line summary of each module from its classes and dependencies, so diagrams show what the code does, not just what it's named. Swap in any OpenAI-compatible provider.
Claude / OpenAI / OllamaGenerate diagrams as Mermaid, D2, or PlantUML. All formats support every level — L0 through L3 — so you can embed diagrams anywhere: GitHub, Confluence, Notion, or your own docs.
Mermaid · D2 · PlantUMLRun archilens serve for a dark-themed SPA. Click any module in the L1 diagram to drill straight into its L2 component view. Search, filter, and re-analyse without leaving the browser.
Point ArchiLens at your semver tags and watch your architecture grow over time. Detects module additions, removals, and LOC trends with a Mermaid timeline diagram and full markdown report.
archilens historyThe diff command compares architecture between any two Git refs without touching the working tree. The GitHub Action (on the Marketplace) posts one architecture-diff comment per pull request and can fail the build on layer-skipping or fan-out rules — no API key needed. With your own key, the comment opens with an AI review of the change. See a real one.
Give Claude Code, Cursor and other MCP clients your architecture: which module owns a file, which capability a feature belongs to, and whether a new import breaks a rule — checked before the agent edits. Re-analyses as the code changes.
archilens serve-mcpReal Mermaid output from a sample microservices repo — no configuration required.
Modules detected automatically from directory structure and import analysis
Modules grouped by the business capabilities you declare, with optional one-line AI summaries
Dependency edges come from real import statements — not guesses
Click any node in the interactive viewer to drill into the next level
Each level zooms in further — start with the big picture and drill into any module.
Your system and its external actors — users, third-party services, databases. The 30,000-foot view.
All detected modules and the dependency edges between them, grouped by capability layer.
Classes, interfaces, and inheritance within a single module. Rendered as a class diagram.
Sequence diagrams for each detected process flow — request handling, async events, and more.
No sign-up. No API key required: without one, ArchiLens runs the full static analysis and skips AI enrichment (or pass --no-ai).
Works with Python 3.10+ on macOS, Linux, and Windows.
Point it at any Git repo — local or freshly cloned.
Open the interactive viewer or read the generated markdown diagrams.
ArchiLens is free, open source, and runs entirely on your machine.
