Supported Languages and Formats

Atlas recognizes programming languages, templates, structured project files, documents, and content formats. Capability depth varies by format: some languages carry a reference-validated call graph, others a structural or content index only. Use this page to determine what Atlas can prove for the languages that matter to you.

Capability Levels

Use these distinctions when evaluating a repository — "indexed" alone does not tell you whether caller and impact queries will be accurate.

Level	Typical capability
Code graph	Symbols, references, calls, and related-code context
Structural index	Named constructs and searchable structure
Content index	Searchable text or extracted document content

Indexing a format does not guarantee identical call-graph or symbol accuracy across every language.

Language Maturity Ladder

Consult the ladder before relying on graph queries (callers, impact, path) for a language: levels reflect validation depth, not just support. An L2 language still indexes and searches; it has not yet reached verified call-graph resolution. Atlas covers 40 code languages across these levels:

Level	Meaning	Languages
L5 — Reference-validated	Call graph cross-checked against an LSP server or SCIP indexer	C, C++, Dart, Fortran, Go, Java, JavaScript, Lua, PHP, Python, Rust, TypeScript, Zig
L4 — Real-repo call graph	Who-calls resolved and proven on a real repository	Apex, Astro, C#, Elixir, ETS, Groovy, Julia, Kotlin, R, Scala, SQL, Svelte, Swift, Verilog, Vue — plus, pending real-repo proof: Bash, Blade, BYOND, EJS, Objective-C, Pascal, PowerShell, Razor, Ruby
L2 — Real-repo tested	Runs on real code; call graph not yet resolved	Delphi, Terraform
L1 — Indexed	Parsed and symbols extracted	P4

The "pending real-repo proof" languages at L4 have resolved call graphs on fixtures but have not yet been proven against a real repository. Treat them as L4 capability with weaker evidence until the proof lands.

Content Formats

Beyond code, Atlas indexes approximately 24 content formats (JSON, YAML, HTML, PDF, and others) for search. These are content and structural indexes, not call graphs:

Templates and frontend: HTML, CSS, Vue, Svelte, Astro, EJS, Razor,

Blade, Markdown, and MDX

Structured and project files: JSON, YAML, TOML, XML, plist, CSV, TSV,

Protocol Buffers, Go module files, .NET project files, Makefiles, Dockerfiles, configuration files, and plain text

Documents and media: PDF, DOCX, XLSX, PPTX, and common image formats,

indexed for content discovery

Treat document and media formats as content indexes rather than programming-language call graphs.

Check Evidence for Your Own Repositories

Fixture compatibility is not a substitute for production-repository accuracy — verify support on the code you actually work with before depending on it.

atlas index .
atlas stats
atlas report --format plain

atlas index . parses symbols, edges, and routes and persists the graph and

lexical index.

atlas stats shows graph and index telemetry statistics for the indexed

repository, including recent snapshot telemetry rows.

atlas report composes the snapshot's graph stats, top hubs, and top

communities; --format plain prints the Markdown report directly.

Low edge counts or missing symbol kinds in stats and report output are the fastest signal that a language sits lower on the ladder than your workflow needs. Evaluate important languages on representative repositories and pin the Atlas version used for the evaluation.

See Indexing and Reindexing for delta versus full reindex behavior and CLI Reference for the full command surface.

Published Compatibility Evidence

Language support changes across releases. The public benchmark site includes a dated, evidence-graded compatibility view with an interactive per-language matrix:

For how those numbers were measured, what they do and do not prove, and how to reproduce them, read Benchmarks and Methodology.