Sessions

Sessions shows each agent’s live terminal. You start and stop agents here, link a terminal to the agent’s MCP connection, and type straight into a running session.

Open it from Sidebar › Agents › Sessions. Clicking a tile on Agents › Team also opens the agent here. An amber dot on Sessions in the sidebar means an agent needs your attention.

Agent list

The list on the left has the heading Agents.

Part	What it does
Collapse button	Collapse to avatars shrinks the list to avatars. Click it again (Expand agent rail) to expand.
Search agents	Filters the list by name.
Online (n)	Agents with a running terminal.
Offline (n)	Agents without a terminal.
Add Agent	Opens the Agent Editor to create an agent.

Each row shows:

The avatar, with a ring for the status and a green link badge when the terminal is linked to the agent.
The name. Another run of the same agent shows a small label with the original agent’s name.
For a running agent: the status (Working, Needs Attention, Idle or Error), the goal it was started in, and User typing when you have unsent text in its terminal.
A muted bell when the agent receives no events.

Hover over a row for its buttons:

Button	Shows when	What it does
Start	The agent is offline	Starts the agent.
Run another	The agent is running	Starts a second run of the same agent and selects it.
Stop	The agent is running	Stops the agent’s terminal.

Drag a row by its handle to change the order. The order is also used by Order sorting on Agents › Team. Dragging is off while you search.

Agent header

Above a running terminal:

Part	What it shows or does
Avatar	The agent’s avatar.
Name menu	The name of the linked agent with a green dot, or No agent linked with a grey dot. Click it for the agent menu. See Name menu.
Status	The agent’s status. Click it to open Session details, with tabs Overview, Signals, Patterns and Raw.
Prompt cache	The time left on the prompt cache, for example `42 min left`.
Context meter	How much of the context is used, for example `62% context`. From 75% it turns amber. Click it to Redeploy instructions.
Recording	Shows while terminal logs are captured.
Backlog button	The number of items in the agent’s backlog. Hover to preview, click to open the backlog in a side panel. You can add, edit, remove and reorder items there.
Notebook button	Hover to preview, click to read the agent’s notebook in a side panel.
Requests	The agent’s open questions and notifications. The badge counts questions. See Activity.
⋮	The terminal menu. See Terminal menu.

Item	What it does
Edit agent…	Opens the Agent Editor.
Run another	Starts a second run of this agent.
Start in goal / Run another in goal	Lists goals that have their own folder. Starts the agent, or another run, in that goal’s folder.
Available Agents	MCP agents that are connected but not linked. Click one to link it to this terminal.
Unlink agent	Removes the link.

When nothing is connected and nothing is linked, the menu says No agents connected.

Item	What it does
Open conversation	Opens the agent’s conversation in Agents › Agent Insights. Shows when Neova has one.
Deploy instructions	Sends the agent’s instructions to the terminal.
Redeploy instructions	Clears the agent’s context, then sends its instructions again.
Capture terminal logs / Stop capturing logs	Records the terminal output to a log file. When you stop, Neova shows the file name.
Restart terminal	Closes the shell and opens a new one. The start command does not run again.
Close terminal	Stops the session after you confirm with Close.

Terminal

The agent’s live terminal. Click in it to type.

When the agent’s backlog has items and the terminal has just started, a banner shows n items waiting, Queued while this terminal was closed. Click Resume to have the agent pull them, Review to see them, or Discard to clear them. Neova asks you to confirm. The × hides the banner until the next session.

Offline agent

When you select an agent that isn’t running, the page shows its overview instead of a terminal:

The name, description, when it was last active and its start command.
Instructions: the first line, with Edit instructions.
Events, Backlog and Notebook cards, with Manage, Open backlog and Open notebook. Each opens that tab in the Agent Editor.
Edit Agent and Start Terminal buttons.

How to start an agent

Select the agent in the list, or hover over its row.
Click Start Terminal, or Start on the row.

Neova opens a terminal, runs the start command, waits for the agent to be ready and sends its instructions.

How to start an agent in a goal’s folder

When a goal has its own folder, the start buttons get an arrow.

Select the offline agent.
Click the arrow next to Start Terminal. The menu says Start in a goal’s own folder.
Choose the goal.

The agent’s row then shows the goal’s title. For a running agent, use Run another in goal in the name menu. See Goals.

How to stop or restart an agent

To stop: hover over the row and click Stop, or choose Close terminal in the ⋮ menu and confirm.
To restart the shell: choose Restart terminal in the ⋮ menu.
To give the agent a clean context with its instructions: choose Redeploy instructions in the ⋮ menu.

How to link a terminal to an agent

A linked terminal receives chat messages and events for that agent. When an agent connects, Neova can link it for you. With Auto-link on name match on, it links when the names match. Neova can also ask Link Agent to Terminal? Click Yes, link agent, or Not now.

To link by hand:

Click the name menu (No agent linked) above the terminal.
Under Available Agents, click the agent.

The dot turns green and the name shows. To undo it, choose Unlink agent.

How to type into a live session

Click in the agent’s terminal.
Type your message and press Enter.

While you type, the agent shows User typing on Team and in the list. To add an image, drag it onto the terminal (Drop image here). Neova saves it and types a reference to it at the cursor, without sending. To add a task, wiki page or other item, press ⌘⇧K. See Insert Reference.

Team and Agent Editor
Settings › Agents for Auto-link on name match and detection profiles
Tools › Terminal for plain shells without an agent
MCP tools agents use for terminals: start_agent, restart_agent and terminal_send