Commands
Type / in the chat and the list narrows to the
commands whose name starts with what you have typed, in alphabetical order.
Enter runs the highlighted one. The commands below are
typed in a session. Skills you installed appear in the same list, marked
skill, and so do saved workflows, marked workflow — a workflow
starts as soon as you run it.
Only some commands take text after the name: /steer,
/goal, /allow, /paths,
/worktree, /mcp, /skills and the skills
themselves. For every other command the rest of the line is ignored:
/model gpt just opens the model window.
Steering a mission
| Command | What it does |
|---|
/steer <text> | Redirect the running mission
without stopping it; it reads the text at its next step. With no mission
running, the text starts one. Picked from the list, it puts
/steer in the input field for you to finish. |
/resume | Continue the last mission that failed or
was cancelled, from where it stopped. Says so if there is none, and is
refused while a mission runs. |
/compact | Summarise the conversation now to free
up context. Refused while a mission runs (it compacts on its own if it grows
too big). |
/verify | Run the project check (tests, build —
whatever this project uses) right now. Esc
cancels it. |
/goal <condition> | Keep working until a
condition is met, checked after every turn. /goal shows it,
/goal clear stops it. Applies to this one mission. See
Setting a goal. |
Looking at changes
| Command | What it does |
|---|
/diff | Everything uncommitted, coloured and
scrollable, in a window. |
/undo | Revert the last file-changing step. Files
that step created stay — /redo brings the change
back. Refused while a mission runs (stop it first with
Esc). |
/redo | Restore the step you last undid. Also
refused while a mission runs. |
/review | Review the changes on this branch from
several angles. Reports, changes nothing. |
/security-review | The same for security: what an
attacker could do with this change. |
Project memory
| Command | What it does |
|---|
/pending | The gate, a full screen with tabs:
facts Nemocode wants to remember, conflicts it found, sync reviews and folders.
Approve or reject each one. |
/curator | Switch the chat to the curator — talk
about what is remembered and why. |
/requests | The curator's requested changes to the
memory: approve or reject them one after another. |
/curators | Every curator chat of this project,
including the harvest sessions — open one to carry on. |
/harvest | Let the curator harvest this session
now. ctrl+t opens its harvest session, where you
watch it and talk to it. |
/bootstrap | Seed the memory from the project's own
README and docs (and a few central code files). The result is proposals that
wait at the gate, not facts. |
/learn | Learn from the project now: reindex the
code, flush the notes that are waiting, and seed from the docs if the memory
is still empty. What comes out waits at the gate. |
/areas | Which topics this session gets preloaded
(Space ticks, a clears
the ticks, Enter saves). None ticked preloads all
of them; whatever is not preloaded stays available on request. |
/pum-folders | Which folders of the project feed the
memory: the gate, on its Folders tab. None ticked means the whole
project. |
/index | Rebuild what Nemocode derives from the code
itself (symbols, structure). |
/re-pum | Rebuild the memory's search index with
the embedding model that is active now. Asks first; the default answer is
Cancel. |
Setup and control
| Command | What it does |
|---|
/connect | Connect a model provider. |
/model | Open the model window for the main agent.
Text after the command is ignored. |
/mode | Open the permission-mode window. |
/effort | How hard the model should think — the
general level. A single role has its own level under Settings ▸ Models
(e). |
/config | The settings window, with tabs Status,
General, Project, Display, Advanced, Models, Usage and Stats. |
/memory | The Agent Memory screen: skills with history and
differences, restore, archive, create and edit by hand, search and filters. |
/allow | The command allowlist for this project;
/allow <command> adds one. Removing one is under
Settings ▸ Project ▸ Allowed commands. |
/paths | Path limits for THIS session:
/paths allow|deny <path>, /paths clear to go
back to the project's own limits. Without an argument it shows what applies.
Takes effect with your next message. |
/permissions | Permission mode, allow/ask/deny
rules, block and confirm lists. |
/worktree | Give this session its own git worktree:
on, merge, off; without an argument it
shows the state. Changing it is refused while a mission runs. |
Bigger runs
| Command | What it does |
|---|
/ultracode | Build, then have it torn apart by an
adversarial review, then fix — at maximum reasoning effort. Puts the keyword
in the input field; you add the task. |
/ultrareview | A multi-angle review that ends in a
report instead of changes. |
/ultrathink | Run just the next mission at maximum
reasoning effort. |
These cost noticeably more than a normal mission —
they run several agents and think longer. Nemocode asks before starting one —
also when the keyword came from one of these commands or from an ordinary
message.
Information
| Command | What it does |
|---|
/usage | Opens the Usage tab of the settings window:
tokens and cost. |
/context | Where the context is going right
now. |
/todo | The full task plan — the pinned version in
the chat is clipped. You can add, rename, delete and re-status items; the
agent reads the change at its next step. |
/history | Browse or search this project's past
sessions. Survives /compact and a new session. |
/ideas | The project's list for later: bugs,
features, notes Nemocode found but did not act on. See
History and ideas. |
/hooks | Which project hooks are active; add or
remove them from there. |
/help | All commands. |
Extending
| Command | What it does |
|---|
/skills | Skills: show, install from a git repo or
folder, remove. |
/mcp | MCP servers: show, add, remove, test. |
/reload | Re-read skills, agents and project files
after you changed them, and rebuild the context from scratch on the next message: the
environment and mode notes are renewed, and everything the project memory told the agent so
far is gathered into one block that shows the current state. The same happens by itself after
the prompt cache has gone cold or the server was restarted. |
/clear | Clear the session's events from the
screen (the session itself keeps them). |
/update | Check for a new release and install
it. |
/nemocode-update | Same as /update. |
/schedule | Scheduled jobs of this project: list,
create, cancel. |
/agents | Custom agents: list, create, delete, use
in this session. |
/connectors | Repos whose issues start missions:
add, poll, remove. |
/licence | Licence status, sign in again, log out
and free this seat. |
/doctor | Health checks of this installation. |
/export | Export all projects to a JSON file. |
From the terminal
Outside the session screen, the nemocode program itself takes a few
commands. nemocode help lists them all.
| Command | What it does |
|---|
nemocode | Start the local server if needed and open
Nemocode. |
nemocode update | Install a newer release;
--check only reports. |
nemocode doctor | Check prerequisites, provider
credentials and whether the server is healthy. |
nemocode logs | Show the last lines of the log;
-f follows it. |
nemocode service install | Run the server as a login
service, so it is already up when you start. |
nemocode integrate claude|codex|opencode|claude-desktop|status | Let another
coding agent use this machine's project memory. See
Other coding agents. |
nemocode export [file] | Write every project's
metadata and memory graph as JSON (to the terminal without a file). |
nemocode logout | Release this machine's seat and
forget the licence. |