Agentlas
Agentlas
Search docs…⌘K
← Back to app
IntroductionFeaturesIntegrationsLegalTips & TricksChangelog
Agentlas terminal

Talk to your agents

You type, your agent answers — word by word. When it does real work (reads a file, edits code, runs a command) you watch it happen, line by line.

Start a chat

agentlas              # pick from your agents
agentlas seo          # open one agent by name
agentlas firm shop    # open a company (the CEO delegates)

What a turn looks like

At the you › line, type what you want and press Enter. The answer streams in. If the agent needs to touch your project, you see a line like ⏺ Read app.ts or ⏺ Bash npm test, then a short result under it.

Slash commands

Type a word starting with a slash to steer the session instead of talking to the agent.

/help
Show every command.
/agents · /agent <name>
List your agents, or switch to one.
/firms · /firm <name>
List companies, or switch to one.
/runtime <name>
Change which AI runs the work (see Runtimes & permissions).
/model <id>
Pick the model (only when using your own API key or Ollama).
/permission <level>
Set what the agent may do: read, write, or full.
/cwd [path]
Show or change the working folder.
/memory
Show the memory being fed to the agent.
/import <path>
Bring in a local folder as an agent or team.
/clear
Wipe the chat and redraw the splash.
/doctor
Check runtimes and data.
/exit
Leave.
TIP · Stop or leave

Press Ctrl-C once to stop the answer it is writing right now. Press Ctrl-C again (or type /exit) to close the terminal.

One-shot, for scripts

You do not have to sit in the chat. Run a single task and get the answer back — handy for scripts and pipes.

agentlas run seo "write 3 title ideas for this page"

# read the task from a pipe
cat notes.md | agentlas run seo
Next: runtimes & permissions →
Who answers, and what they are allowed to touch.
Talk to your agents · Agentlas docs