CLI reference
The glaux command drives everything the dashboard can do. It talks to the running
server over HTTP, so it works locally or against a remote instance (--host / --port).
Run glaux --help, or glaux <command> --help, for the authoritative list.
Server
| Command |
Does |
glaux up [--foreground] [--open] |
Start the server as a background daemon |
glaux down |
Stop the daemon |
glaux restart |
Stop, wait for the port to free, start again |
glaux rebuild [--no-build] |
Rebuild the dashboard, then restart |
glaux status |
Model, safety, ledger size, capabilities, cache |
glaux serve |
Run in the foreground |
glaux open |
Open the dashboard in a browser |
glaux desktop [--build] |
Launch the native desktop app |
Driving the swarm
| Command |
Does |
glaux run "…" |
Dispatch a directive; streams the run live |
glaux run "…" --json |
Machine-readable result (for scripts) |
glaux chat |
Interactive terminal session |
glaux search "…" |
Search the web via the agent's tools |
glaux remember "…" [--layer] |
Store a fact in memory |
glaux memory <layer> [query] |
Browse semantic / episodic / procedural memory |
Directives can also be piped: echo "summarize this" | glaux run.
| Command |
Does |
glaux verify |
Verify the signed ledger chain |
glaux verify --file PATH |
Verify a ledger offline — no server, no trust in the app |
glaux metrics |
Tasks, tokens, cost, skills synthesized |
glaux eval |
Run the benchmark suite |
glaux agents |
The swarm's signed agent identities |
Skills
| Command |
Does |
glaux skills |
List learned skills (📌 marks promoted permanent tools) |
glaux skills synth "…" |
Synthesize a new skill |
glaux skills approve <id> |
Approve a pending skill (gated mode) |
glaux skills reject <id> |
Reject it |
Capabilities
| Command |
Does |
glaux workspace [on\|off] |
The jailed file workspace; lists files + tools |
glaux playbooks [list\|add\|rm\|on\|off] |
Authored markdown procedures |
glaux subagents [list\|add\|rm] |
Custom specialists in the router's roster |
glaux cache [status\|on\|off\|clear] |
Semantic answer cache |
glaux personas [list\|add\|rm] |
Saved system prompts |
glaux recipes [list\|add\|rm] |
Saved task templates |
Examples:
glaux playbooks add "Deploy checklist" --desc "ship safely" --body "1. run tests…"
glaux subagents add "Legal reviewer" --prompt "You review contracts…" --iters 6
Self-evolving harness
| Command |
Does |
glaux evolution |
Routing accuracy, learned config, pinned tools, retired roles |
glaux evolve |
Run one evolution pass (all operators, each gated) |
glaux evolution revert |
Roll back the most recent change |
glaux evolution auto on --every 20 |
Evolve automatically every N episodes |
glaux evolution unpin <skill_id> |
Un-promote a permanent tool |
glaux evolution unretire <role> |
Bring a retired specialist back |
Safety
| Command |
Does |
glaux safety [live\|dry-run\|halted] |
The kill switch |
glaux approval [auto\|gated] |
Whether a human approves synthesized skills |
Models and account
| Command |
Does |
glaux models [--set BACKEND[:MODEL]] |
Show or switch the LLM |
glaux login / logout / whoami |
Account session |
glaux messaging |
Messaging gateways (admin) |
glaux version |
Version |