-
Notifications
You must be signed in to change notification settings - Fork 0
02 CLI cli reference
The Krnl-AI CLI (krnlai) is the primary interface for interacting with the community runtime.
Start an interactive TUI chat session with the local embedded kernel.
krnlai chat --local
krnlai chat --local --model llama3.1Create a new agent project with scaffold files.
krnlai init my-agentRun the agent once with the given input text.
krnlai run "analyze this dataset"
krnlai run --interactiveStart the sidecar HTTP server for editor integrations and automation.
krnlai serve --local --port 5117Search semantic memory for relevant documents.
krnlai memory search "project decision"Create a snapshot of current memory state.
krnlai memory snapshotView memory usage statistics.
krnlai memory metricsExecute safety checks against the current configuration.
krnlai safety runDisplay the status of all safety layers.
krnlai safety statusRun a full security audit of the safety system.
krnlai security auditPerformance benchmark the safety system (default: 1000 iterations).
krnlai security benchmark 5000Generate an HTML security report.
krnlai security report report.htmlList installed skills.
krnlai skill listExport a skill for sharing.
krnlai skill export my-skill skill.jsonImport a skill from a file.
krnlai skill import skill.jsonList learned policies.
krnlai policy listShow details of a specific policy.
krnlai policy show policy-1List all sessions.
krnlai session listExport a session for sharing or analysis.
krnlai session export session-1 session.jsonShow all configuration values.
krnlai config listSet a configuration value.
krnlai config set model llama3.1Show the current status of the Krnl-AI runtime.
krnlai statusCheck the health of the runtime and configured providers.
krnlai healthUpgrade the CLI to the latest version.
krnlai upgradeList available project templates.
krnlai templates listCreate a new project from a template.
krnlai new agent my-agent
krnlai new tool my-tool
krnlai new policy my-policyKrnl-AI Community — MIT License