Quick Start
This page covers installing Kappa Graph on a server and connecting client tools to it. Choose your path:
| Goal | Section |
|---|---|
| Install the platform on a Linux server | Platform Installation |
| Install the CLI or MCP client | Client Tools |
| Set up a development environment | Development Setup |
Platform Installation
One command installs Kappa Graph on a Linux server:
The interactive installer:
- Installs Docker if not present
- Downloads container images from GHCR
- Generates infrastructure secrets
- Configures TLS (optional)
- Creates the admin user
Headless installation
For scripted or CI deployments, pass flags directly:
curl -fsSL https://raw.githubusercontent.com/aaronsb/knowledge-graph-system/main/install.sh | bash -s -- \
--hostname kg.example.com \
--ssl letsencrypt \
--ssl-email admin@example.com \
--ai-provider openai \
--ai-key "$OPENAI_API_KEY"
To configure AI credentials after installation, omit the AI flags:
curl -fsSL https://raw.githubusercontent.com/aaronsb/knowledge-graph-system/main/install.sh | bash -s -- \
--hostname 192.168.1.100 \
--skip-ai
Installer flags:
| Flag | Values | Notes |
|---|---|---|
--hostname |
hostname or IP | Required for headless mode |
--ssl |
offload, selfsigned, letsencrypt, manual |
Defaults to self-signed |
--ssl-email |
email address | Required for letsencrypt |
--ai-provider |
openai, anthropic, ollama |
|
--ai-key |
API key string | |
--skip-ai |
— | Configure AI provider after install |
--gpu |
auto, nvidia, amd, cpu |
|
--admin-password |
password string | Auto-generated if omitted |
Verify the installation
After the installer completes, check that all containers are healthy:
Expected output:
Container Status:
knowledge-graph-postgres Up (healthy)
knowledge-graph-garage Up (healthy)
kg-api Up (healthy)
kg-web Up (healthy)
kg-operator Up
In production overlay mode (--container-prefix=kg) the infrastructure containers are named kg-postgres and kg-garage. In development mode, the application containers are kg-api-dev and kg-web-dev.
Access points after a successful install:
- Web UI:
https://your-hostname(orhttp://localhost:3000in dev mode) - API docs:
https://your-hostname/api/docs - Admin username:
admin(password shown at end of install, or set via--admin-password)
Managing the Platform
operator.sh is the primary management tool after installation.
# Daily operations
./operator.sh start # Start all containers
./operator.sh stop # Stop all containers
./operator.sh status # Show container health
./operator.sh restart api # Restart one service
# Logs
./operator.sh logs api # Last 100 lines of API logs
./operator.sh logs api -f # Follow API logs
./operator.sh logs web # Web container logs
# Updates
./operator.sh update # Pull latest images (no restart)
./operator.sh update operator # Pull a specific image
./operator.sh upgrade # Pull, migrate, and restart
# Maintenance
./operator.sh teardown # Remove containers (data volumes kept)
./operator.sh teardown --full # Remove containers and volumes
Configuration shortcuts:
./operator.sh shell # Open configuration shell (configure.py)
./operator.sh admin # Manage the admin user
./operator.sh ai-provider openai # Switch AI provider
./operator.sh api-key openai # Store a provider API key
./operator.sh embedding # Configure embeddings
./operator.sh query 'SQL' # Run a SQL query against PostgreSQL
Inside the configuration shell (./operator.sh shell):
configure.py status # Show current configuration
configure.py ai-provider openai # Switch AI provider
configure.py api-key openai # Store a provider API key
configure.py oauth --list # List OAuth clients
Development Setup
Clone the repository and run the init wizard:
git clone https://github.com/aaronsb/knowledge-graph-system.git
cd knowledge-graph-system
./operator.sh init
The wizard:
- Generates infrastructure secrets (
.env) - Detects your GPU (NVIDIA, AMD, or CPU)
- Starts all containers
- Prompts for admin password and AI provider
For non-interactive development setup:
Do not edit .env manually — it is generated by ./operator.sh init and managed by the operator.
Client Tools
Kappa Graph ships three client tools: the kg CLI, the MCP server (bundled with the CLI), and the optional FUSE filesystem driver.
All-in-one installer
The client installer detects your OS, installs prerequisites, and configures authentication:
curl -fsSL https://raw.githubusercontent.com/aaronsb/knowledge-graph-system/main/client-manager.sh | bash
It prompts for the API URL and credentials during setup.
To upgrade or remove later:
Manual installation
kg CLI (requires Node.js ≥ 20.12.0):
FUSE driver (optional):
Manual upgrade:
Configure the CLI
After installation, point the CLI at your Kappa Graph instance and authenticate:
Configuration is stored at ~/.config/kg/config.json. The FUSE driver reads credentials from the same file.
CLI quick reference
kg health # Check API connection
kg search "query" # Search concepts
kg ingest file.pdf # Ingest a document
kg --help # Full command list
MCP server
The MCP server is included with the CLI. To configure it for Claude Desktop:
kg mcp-config # Generate Claude Desktop config snippet
kg oauth create-mcp # Create an OAuth client for the MCP server
FUSE filesystem
kg-fuse init ~/Knowledge # First-time setup
kg-fuse mount # Start all configured mounts
ls ~/Knowledge/ # Browse ontologies
kg-fuse status # Check mount status
See Use the FUSE Drive for full setup details.
Troubleshooting
kg: command not found after installation
Ensure ~/.local/bin is in your PATH:
Add this line to ~/.bashrc or ~/.zshrc for persistence.
Authentication errors
Wrong API URL
Platform containers not starting
For deeper platform issues, see Troubleshooting.
What's Next?
- Production Deployment — TLS, scaling, monitoring
- Configuration Reference — All settings
- Upgrading — Version updates
- Troubleshooting — Diagnostics and fixes