Code Intelligence

Beyond scanning, xgrep is a fast, AST-powered way to navigate and understand a codebase — for humans and AI agents alike. It indexes symbols and call relationships into a code graph and a search index (cached in .xgrep/, rebuilt incrementally), so answers are precise (real symbols with locations) and navigable, not just text hits.

Search is lexical and structural, not semantic — you find code by a name or a string it contains, then traverse call relationships. When you only know what code does, not its name, start with an overview to learn the vocabulary, then search those names and follow the references.

Two entry points, both with --json on every command:

• xgrep inspect — overview, symbol/text search, go-to-definition, hover, references, implementations, file outline, change-impact, and dependencies.
• xgrep graph — build and query the call graph: callers, callees, call paths, reachability, and N-hop neighborhoods with inlined source.

The examples on this page and in the command reference run against expressjs/express — clone it and follow along:

git clone https://github.com/expressjs/express && cd express

What you can do

Understand a new codebase

Start with a one-shot map — languages, packages, entry points, key types — then drill into a file or a name:

xgrep inspect overview .                  # high-level map of the whole tree
xgrep inspect outline lib/router/route.js # every symbol in one file, in order
xgrep inspect symbol "Router"             # find a symbol by name (exact → prefix → substring)

Find code fast

Sub-millisecond trigram search after the first index, with regex, language, and glob filters:

xgrep inspect search "res.send" --file-pattern "*.js"
xgrep inspect search "req\.(query|body|params)" --regex --lang javascript

Navigate like an IDE

Resolve, explain, and cross-reference a symbol at a position or by name:

xgrep inspect definition --file lib/router/route.js --line 43  # what is this?
xgrep inspect hover --file lib/router/route.js --line 43        # docs, type, signature
xgrep inspect references "matchLayer"                           # callers + callees
xgrep inspect implementations "Router"                         # who implements this interface

Assess change impact

Before you touch a function, see its blast radius — direct and transitive callers, affected files, and a risk score:

xgrep inspect impact "matchLayer"        # blast radius + risk score
xgrep inspect deps "matchLayer" --depth 3 # up/downstream call dependencies
xgrep graph reachable "acceptParams"     # everything this function can reach

Triage a finding

Pull a finding's call neighborhood with the source inlined — topology and code in one shot, ideal for deciding whether a finding is reachable:

xgrep graph context "matchLayer" --depth 2

Power an AI agent

Every command emits --json, and the same capabilities are exposed as MCP tools via xgrep mcp, so an agent can map, navigate, and triage a codebase without bespoke tooling. See AI agents for end-to-end workflows that combine code intelligence with scanning.

Reference

• Inspect — every inspect command, its flags, caching behavior, the MCP tool mapping, and supported languages.
• Code graph — building and querying the call graph.