Getting started

Install

Find available packages on GitHub Releases. The release workflow builds an Apple silicon macOS DMG and a 64-bit Windows ZIP. Windows packages are unsigned; the latest engine upgrade was validated on macOS, not Windows.

On macOS, open the DMG and copy the app into Applications. On Windows, extract the entire ZIP, preserving the plugins folder beside the executable.

The Command line & AI button (the robot icon) in the app title bar installs the khonsu command on PATH and can add the agent skill to Claude Code. The app executable also accepts CLI subcommands directly.

khonsu --version

Create a small graph

Download hosts.csv and dependencies.csv into the same working folder. They describe three hosts and two dependencies.

Run these commands in that folder. Keep the graph closed in the desktop app during the import.

khonsu init graphs/demo
khonsu source add --db graphs/demo --plugin csv --name demo \
  -c node_file=hosts.csv -c node_label=Host -c node_id_column=id \
  -c edge_file=dependencies.csv -c edge_type=DEPENDS_ON \
  -c edge_source_column=source -c edge_target_column=target --test
khonsu plugin sync --source demo --dry-run
khonsu plugin sync --source demo

The first sync previews counts and samples without writing; the second imports the graph. source add stores the CSV paths as absolute paths, so later syncs work from any folder.

Query and explore

khonsu schema --db graphs/demo
khonsu query --db graphs/demo --cypher \
  "MATCH (a:Host)-[r:DEPENDS_ON]->(b:Host) RETURN a, r, b"
khonsu open graphs/demo

In the desktop query editor, run the same Cypher query. Returning whole nodes and relationships gives the graph canvas the information it needs to show connections.

Continue with queries or your first dashboard.

Build from source

From a checkout of the app repository:

cargo build --workspace
cargo run -p khonsugraph

The app uses Rust, GPUI and HawDB. See the repository README for release packaging and platform-specific setup.