The graphlens CLI

graphlens-cli installs the graphlens command — a Typer app with five subcommands. This page is a practical tour; the CLI API reference lists every flag and default.

pip install "graphlens-cli[all]"
graphlens --help

Command	Purpose
analyze	Parse a project; print stats or serialize the graph to JSON
query	Run callers/callees/references/neighbors on a saved graph
visualize	Build an interactive HTML graph viewer
neo4j	Export the graph to a Neo4j database
mcp	Serve a saved graph to agents over the Model Context Protocol

Selecting languages

analyze, visualize, and neo4j accept a --lang option. The default, auto, asks the registry for every installed adapter and keeps the ones whose can_handle() returns true for your project. You can also name adapters explicitly, comma-separated, to analyze a polyglot repository in one pass:

graphlens analyze ./monorepo --lang python,typescript

analyze

# Print node/relation statistics
graphlens analyze ./my-project

# Serialize the graph to JSON (the indexing step for CI and agents)
graphlens analyze ./my-project --output graph.json
graphlens analyze ./my-project --format json        # JSON to stdout

# Fail the command if the resolver did not complete
graphlens analyze ./my-project --strict

Option	Default	Description
--lang	auto	Adapter(s) to use, comma-separated
--format, -f	text	text (stats) or json (serialized graph)
--output, -o	—	Write the serialized JSON graph to this path
--strict	off	Exit non-zero if the resolver status is not ok

--strict is the flag that makes graphlens safe to gate a pipeline on — see CI integration.

query

Operate on a graph you already saved with analyze --output:

graphlens query process_order --graph graph.json --op callers
graphlens query process_order --graph graph.json --op callees
graphlens query OrderService.save --graph graph.json --op references
graphlens query OrderService.save --graph graph.json --op neighbors --depth 2

Option	Default	Description
node (argument)	—	Node id, qualified name, or short name
--graph, -g	—	Path to a graph JSON file (required)
--op	callers	callers | callees | references | neighbors
--depth	1	Hop depth for the neighbors operation

The node argument is resolved by id first, then by qualified or short name, so you can pass whichever you have on hand.

visualize

Produce a self-contained HTML file (powered by vis.js) and open it in your browser:

graphlens visualize ./my-project
graphlens visualize ./my-project --show-external --max-nodes 500
graphlens visualize . --output graph.html --no-open

Option	Default	Description
--lang	auto	Adapter(s) to use
--output, -o	graph-<name>.html	Output HTML file
--no-open	off	Do not open the browser automatically
--show-external	off	Include stdlib / third-party external symbol nodes
--show-structure	off	Include CONTAINS / DECLARES structural edges
--max-nodes	1500	Prune low-degree nodes above this count

See the dedicated visualization guide for the viewer's interactions (search, filters, Show callers focus mode).

neo4j

Export straight into a Neo4j database with UNWIND … MERGE Cypher (no APOC required). Install the exporter dependency first:

pip install "graphlens-cli[neo4j]"

graphlens neo4j ./my-project \
  --uri bolt://localhost:7687 --user neo4j --password secret

graphlens neo4j . --wipe --batch-size 200

Option	Default	Description
--lang	auto	Adapter(s) to use
--uri	bolt://localhost:7687	Neo4j Bolt URI
--user	neo4j	Neo4j username
--password	password	Neo4j password
--wipe / --no-wipe	--no-wipe	Wipe existing :Code nodes first
--batch-size	500	Items per Cypher batch

See the Neo4j guide for the resulting schema and example queries.

mcp

Serve a saved graph to LLM agents over the Model Context Protocol. Install the MCP dependency first:

pip install "graphlens-cli[mcp]"
graphlens mcp --graph graph.json

Option	Default	Description
--graph, -g	—	Path to a graph JSON file (required)

See the MCP server guide for the exposed tools and how to wire it into an agent client.

Exit codes

Commands exit non-zero on errors (bad arguments, missing adapter, unreadable graph). analyze --strict additionally exits non-zero when the resolver status is not ok, which is what lets you fail a CI job on a degraded graph.