Give your agent a map
before it explores.

A package or directory — the coarsest grouping unit.

A function, method, struct, interface, type, class, or constant.

An HTTP or API endpoint — method, path, and handler.

A database table or data store (e.g. ActiveRecord model, JPA entity).

An import relationship to an external or internal package.

A whole repository — used when analyzing multiple repos together.

A module declares a symbol, route, or storage entity.

One module imports another.

A function or method invokes another.

A type implements an interface or mixes in a module.

A general dependency edge between any two nodes.

A function creates an instance of a type.

A DI framework provides a dependency (Hilt, Spring, constructor injection).

A class or struct owns a method — synthetic edge for traversal completeness.

Parses the codebase and produces a structured architectural graph. Auto-detects Go, TypeScript, Kotlin, Ruby, Python, Swift, Rust, C, C++, Java, PHP, and OpenAPI from project markers — no per-language config needed. Extracts module, symbol, dependency, route, and relation facts. Pass append: true to merge multiple repositories into a single traversable graph. Pass fresh: true when switching to a different project in the same session — required to avoid silently reusing stale graph state.

Precision filter over extracted facts. Best for targeted lookups when you know what you want. Filter by kind (module, symbol, function, class, route, dependency), name (substring), file, relation, prop, or repo. Output modes: compact (token-efficient), full (all fields), names (identifiers only). Use explore instead when you want a broad view of an entity.

Primary post-snapshot navigation tool. Pass any module name, file path, symbol, or directory and get a structured markdown overview of its contents, dependencies, relations, and outbound HTTP and gRPC client calls (Retrofit, URLSession, fetch, Net::HTTP, http.Client, grpc-go, connect-go, grpc-web, connect-es, Python gRPC stubs). Accepts partial names — path normalization handles trailing slashes and relative paths. Reach for this first; use query_facts for structured filtering or traverse for deep graph walking.

Return source code for a named symbol. Tries exact match first, falls back to substring. Default context window: ~15 lines before, ~45 lines after the definition. Override with context_lines_before and context_lines_after. Works across multi-repo snapshots.

Walk the dependency or call graph from a starting node. direction='forward': what does X import or depend on? direction='reverse': what depends on X? Starting node accepts substring — multiple matches are grouped by match. Filter by relation kind (e.g. imports, calls, implements) or node kind (e.g. module, function). Default depth: 3.

BFS shortest path between two graph nodes. Both from and to accept substring names — if multiple nodes match, the closest pair is used. Answers "how does X reach Y?" or "what is the call chain from main to this handler?". Returns an ordered list of node → relation → node steps.

Return the blast radius of a change — all nodes that transitively depend on the target, grouped by depth. Set include_forward: true to also return what the target depends on (reverse + forward in one call). Target accepts substring matching. Use before refactoring to understand risk surface.

Retrieve computed explainer findings — coverage gaps, unused routes, architectural smells, complexity outliers — filtered by source, repo, or confidence threshold. Use after generate_snapshot to surface what the insight layer found without reading the full insights.json.

Return a cross-repo edge resolution summary — which outbound calls were matched to handlers in other repositories and which remain unresolved. Useful for verifying that multi-repo graphs are correctly linked before running impact analysis.

Pin the current snapshot as a baseline. The baseline is stored in .enola/baseline/ and persists across re-snapshots. Use before a major refactor to enable before/after comparison.

Compare the current snapshot against the pinned baseline and return a structural delta — nodes added or removed, edges changed, new cycles or layer violations introduced. Useful for verifying that a refactor produced only the intended structural changes. Guarded by a comparability check — mismatched configs or extractor versions are flagged instead of silently diffed.

Return the receipt.json for the current snapshot — provenance metadata (enola version, git ref, content fingerprint, config hash, output hash) plus extraction-quality metrics. Use to verify exactly what a snapshot was built from before trusting its output downstream.

Compare two snapshot receipts and report whether they're comparable — same config hash, extractor versions, and repo set. Run before diff_snapshot to catch mismatches that would otherwise make a structural diff misleading.

go/ast

go.mod

tree-sitter AST parser

pom.xml, build.gradle, or .java files (Spring routes / JPA / Lombok DI / Dubbo SPI aware)

tree-sitter

tsconfig.json or package.json — processes .ts, .tsx, .js, .jsx (Next.js & monorepo aware)

tree-sitter (TS extractor)

package.json with vue or nuxt dependency — SFC script block extraction; Nuxt file-based routes inferred from pages/

tree-sitter (TS extractor)

package.json with svelte dependency — .svelte SFC extraction; $lib alias resolved to src/lib/; SvelteKit file-based routes inferred

tree-sitter AST parser

pyproject.toml, setup.py, requirements.txt, or similar (FastAPI / Django / SQLAlchemy / gRPC aware)

tree-sitter AST parser

build.gradle.kts, build.gradle, or pom.xml (Compose / Hilt / Room aware; falls back to a src/main/kotlin source root)

tree-sitter AST parser

Package.swift, .xcodeproj, or .xcworkspace (SwiftUI / UIKit aware)

tree-sitter AST parser

Gemfile (Rails / ActiveRecord / Packwerk aware), or extensionless scripts with a Ruby shebang for Gemfile-less repos

tree-sitter AST parser

.c / .h files

tree-sitter AST parser

CMakeLists.txt, .cpp / .hpp / .cc files (header/source merging, namespaces, templates)

tree-sitter AST parser

composer.json or .php files (WordPress / Laravel / Symfony aware)

tree-sitter AST parser

Cargo.toml (cross-crate and submodule resolution for Cargo workspaces, Axum route detection)

YAML/JSON scanner

Any file containing openapi: or swagger:

Repository root path

"."

Glob patterns for files/dirs to skip

vendor, node_modules, .git, tests, build dirs, docs

Enabled extractors

All supported

cycles, layers, cross-repo, coverage, unused-routes, god-class, hotspots, dependency-depth, exported-surface, complexity-outliers

All

Output directory for generated artifacts

".enola"

Token budget for the LLM context summary

16000

Reuse content-hash cache to skip unchanged files on re-snapshot

true