ncode Desktop The workspace
Watching agents work
Every run in ncode is a small team: one top agent, the agents it starts, and every tool call each of them makes. The agents pane on the right of the window shows all of it while it happens, so you always know what is going on and what it costs.
Agent cards#
The Agents view shows one card per agent: its name and task, its status (running, waiting for you, done, failed, stopped), a progress bar, the model it uses, the tokens it has used and what they cost. Under each agent, its tool calls appear as rows: the file it read, the command it ran, the search it made.
- Tree shows who started whom: the lead at the top, its sub-agents below it, their own sub-agents below them.
- Grid lays the same cards side by side, which is easier to scan when many agents run at once.
- Full cards show everything; Compact cards show one line per agent.
Everything that waits for you (an approval, a question) is marked on its agent and in the transcript.
The timeline#
The Timeline view puts every step of every agent on a time axis, so you can see what ran in parallel, what waited and what took long. Click a step to open the inspector: the step whole, with its full input and output.
Changes#
The Changes view lists the files the conversation's agents changed, with a diff for each. From here you can restore a single file, or all the files of a turn, to how they were before (see Rewind). Agents that worked in an isolated copy of the project show their work under Branches until it is merged (see Limits and isolation).
Pause, continue, stop, resume#
Each run's card has the controls that fit its state:
| Button | When | What it does |
|---|---|---|
| Pause | running | holds every agent before its next step; nothing is killed |
| Continue | paused | lets the held agents take their next step |
| Stop | running or paused | stops the run and every agent in it |
| Resume | stopped, failed or interrupted | starts a new run that continues from where the old one ended |
A resumed swarm tells its new lead what the old workers finished, and a resumed consensus run carries on with its rounds. /stop stops everything in the conversation; /resume resumes its last stopped run.
Commands left running#
A command an agent starts may keep running after the step ends: a dev server, a file watcher, a long build. Such commands appear in a Still running strip at the top of the agents pane, with their process id. Click Stop to end one together with everything it started.
Talking to a running run#
Use ↠ Steer on a running run's card to add a message to it: its top agent reads it before its next step. Use ↩ Reply to send a follow-up about a run that has finished.