// docs
Agent Workspace
The S.E.N.T.R.I tab is where your agents live: one main chat, as many agents beside it as the job needs, and every session saved and restored when ServerEngine starts.
Two views
The tab has two views. The main chat is your ongoing conversation with S.E.N.T.R.I. The Agent Workspace is a grid of further agents, each working on its own task at the same time. All of them draw on the same skills and the same memory system.
The main chat
Type into the prompt line, or just talk to it — see Voice. Messages you send from WhatsApp land in this same conversation. Everything that shapes the agent sits in the chat's header and prompt line:
| Control | What it does |
|---|---|
| Scope | Which hosts the agent may act on. It starts on LOCAL (This Computer); on Commander your server groups appear here too. |
| Working folder | The folder icon. It is the boundary for the Coder skill, and the folder your memories are filed under. |
| Model | Pick from your starred Favorites or browse the whole catalog. The chip in the prompt line shows the model and its reasoning effort. |
| Reasoning | Off, Low, Medium, High or xHigh, set per chat from the model chip. |
| New chat | Starts a fresh conversation and resets the context. |
| Microphone and speaker | Record a prompt, and mute or unmute the spoken reply. |
| Budget | For OpenRouter: your spend against the monthly cap, and how long until it resets. |
| Available Skills | The skills you have enabled, listed beside the chat. Hover one to see what it does and every sub-skill it offers. |
Agent Workspace (ADE)
The Agent Workspace is ServerEngine's agent development environment. Every pane is a full S.E.N.T.R.I session with its own scope, working folder, model and reasoning effort, so a job can be split across hosts or projects and run in parallel instead of one chat working through them in turn.
- Press + Add agent to open another pane. Panes sit three to a row, and the split bars between them can be dragged.
- Each pane has its own prompt line and microphone, a New chat button and a close button.
- Give each agent the model its part needs: a quick lookup does not need the same model as a refactor, and every pane can run a different one.
- Slack and Discord channels are answered by agents here as well, one per channel — see Chat Integrations.
Sessions that survive a restart
Every session — the main chat and every agent beside it — is saved and restored with its context when ServerEngine starts, so closing the app does not cost you a conversation. You can switch between workspaces without losing any of them, and change model in the middle of a conversation without losing its history.
Plan mode, checklists and questions
Plan mode
With plan mode on, S.E.N.T.R.I looks before it acts. It can read your code with the Coder skill and ask you questions, but every other skill waits. What it hands back is a plan card: the plan itself, and the phases it will work through. Press Execute to let it start, or Keep planning to refine the plan first.
Task checklist
Any job of more than two steps — or several separate jobs in one request — gets a checklist before the work begins. One phase is in progress at a time, and each is ticked off the moment it is done, so you can tell at a glance how far along the agent is.
Questions
When a decision is genuinely yours — which of two approaches, which file, which trade-off — the agent asks it as a short list of options, each with a line on what choosing it means. An Other answer is always there if none of them fits. Where there is a sensible default, the agent takes it and tells you which one it took instead of asking.
Shortcuts and commands
| Key or command | What it does |
|---|---|
| Enter | Send the prompt. |
| Esc | Interrupt a prompt that is running. |
| Shift+Space | Push to talk. |
| / then Tab | Type / and the start of a skill or host name, then press Tab to complete it. |
| /newchat | Start a new chat and reset the context. |
| /scope | Change the scope. |
Images can be pasted straight into the prompt line or dragged into the chat, with a line saying what you want to know about them. An auto-scroll toggle keeps the transcript following the agent, or holds it still while you read back.
History and audit
Every conversation is kept. The Logs tab has its own S.E.N.T.R.I history view you can search, and unattended runs — scheduled or started from the API — are written there in full under their job ID. Where it all lives is covered in Chat History.
Which plan