Reference
Commands and flags
Launch§
lampson # terminal, on the current folder
lampson --web # browser, http://127.0.0.1:8080
lampson --workspace ~/other # a project without cd
lampson --agent plan # start read-only
lampson --yolo | --strict | --ask
lampson --hub start|stop|status|logs|restart # Lampson as a whole: /docs/workspaces
lampson --daemon start # the resident process for scheduled tasks
lampson --install # keep the hub alive across logins · --uninstall undoes it
lampson --domain your.domain.com --email you@mail # HTTPS on your own domain on a VPS (optional) · --domain off
lampson --pair [https://your.domain.com] # QR in the terminal to pair a phone (VPS over SSH)
lampson --update # git install: pull · npm: tells you the command
lampson --help
Lampson as an engine§
lampson run [--agent ID] [--thread T | --session ID] [--json] "message" # headless: /docs/headless
lampson run --stdio # one open process, one JSON request per line (SDK local mode)
lampson key add [--agent ID[,ID]] [--name N] [--rate N] [--callback URL] # API keys: /docs/api
lampson key list · lampson key remove <id>
lampson channel list · channel on|off <channel> · channel setup <channel> # channels: /docs/channels
lampson plugin list · plugin on|off <name> # plugins without a UI: /docs/plugins
lampson tools list · tools approve [name] # the project's .syn tools: /docs/sdk#in-synsema-with-guarantees
lampson mcp serve [--agent ID[,ID]] [--workspace PATH] # Lampson as an MCP server: /docs/mcp-server
lampson objective add "…" --agent ID | --team ID [--budget N] [--round S] # objectives: /docs/objectives
lampson objective list · show <id> · reply <id> "…" [--approve|--deny] · cancel <id>
lampson team list # teams: /docs/teams
None of these start the hub: they prepare the project folder and do their thing.
Slash commands§
| command | ||||
|---|---|---|---|---|
/model [provider] [model] (also /provider, /setup) | provider → model (listed from the API) → key if missing, with arrow keys; saved for next time | |||
/agent <name> | switch agent: the built-in ones (build, plan, review, explore) and yours; with its model, if it has its own | |||
/ask · /yolo · /strict | permission mode | |||
/new · /sessions · /resume <id> · /delete <id> | sessions | |||
/trace [n] · /tokens · /out [n] · /verbose | what happened, how much it cost, full tool outputs | |||
/files · /memory [name] · /todo | the project tree, the agent's notes, the session's todo list | |||
/procs · /logs <name> · /stop <name> · /kill <pid> | servers the agent started | |||
| `/plugins [on | off | run | remove] …` | plugins |
/computer-use [on|off] | computer use: the agent drives your desktop and browser; alone, the driver's status | |||
/judge [off|jev|laya] | the decision judge; asks for the API key or the folder when missing | |||
/mcp [add|remove] … · /lsp [add|remove] … | MCP and language servers | |||
/skills | skills the agent can see | |||
/agents | sub-agents | |||
/schedule [add|list|remove] … · /approve | scheduled tasks and pending approvals | |||
/image <path> · /paste | attach an image to the next message | |||
/config · /flags | current settings | |||
/update · /help · /exit |
!command runs a command yourself in a real terminal; its output enters the conversation.
Keys§
| key | |
|---|---|
/ | command menu (recent first, filter as you type) |
Tab | complete the command, then its arguments |
↑ ↓ | history, or move in a menu |
Alt+Enter | new line |
Ctrl+O | last tool result in full |
Ctrl+U / Ctrl+W | clear line / word |
Esc | close the menu |
p / d (or arrows) | allow / deny an approval |