Slash Commands

This page covers every command available in Huginn’s / picker — what each one does, when to use it, and real examples.

How the picker works

Type / in the chat input to open the inline picker. It filters in real time as you keep typing — /pl narrows to /plan before you finish.

Built-in commands and user-installed skills share the same list. There is no visual distinction between them.

Workflow commands

These commands control how the agent approaches your request.

/plan

Switches the active slot to the planner model (Chris by default) and primes it for structured output. Use this before writing any code.

/plan
describe a Redis cache layer for the user service

After planning, hand off to Steve with /code.

/code

Switches to the coder model (Steve by default). No planning phase — sends your request straight to implementation.

/code fix the nil pointer dereference in auth.go:47

/reason

Switches to the reasoner model (Mark by default). Slower, more thorough. Use it when you need careful analysis before acting.

/reason should I use a message queue or direct HTTP calls here?

/iterate [N]

Runs the agent multiple times on the same task. Each pass receives the output of the previous one as context. Prompts for N if you don’t supply it.

/iterate 3

Good for code quality, prose polish, or test generation where one pass isn’t enough.

/parallel <task1> | <task2> | ...

Runs multiple tasks at the same time using the default agent. Separate tasks with |. Results appear sequentially, labeled by number, when all complete.

/parallel Write unit tests for auth.go | Write unit tests for session.go | Write unit tests for relay.go

All tasks run concurrently. Use this when they don’t depend on each other.

/swarm <agent>:<prompt> | <agent>:<prompt> | ...

Like /parallel, but you route each task to a specific named agent. Agents run concurrently and stream results into the swarm view.

/swarm Steve:implement the login handler | Mark:review the auth logic for edge cases

Each segment is agentName:prompt. Without a colon, the whole segment is treated as both the agent name and the prompt. Use /swarm with no arguments to re-open the swarm view if you’ve navigated away.

/parallel vs /swarm: /parallel sends all tasks to the default agent. /swarm routes each task to a specific named agent.

Agent & model commands

/agents

With no arguments, shows the agent roster and opens the agents management screen.

Sub-commands:

/agents create Petra claude-opus-4-6
/agents swap Steve claude-sonnet-4-6
/agents rename Chris Architect
/agents persona Mark
/agents delete Petra

Changes are written to ~/.huginn/agents/ immediately and take effect in the current session.

/switch-model

Opens an interactive prompt to change the model for the current session slot without saving it to your config. Takes free-form input.

/switch-model
use claude-opus-4-6 for planning

Use this for a one-off override without touching your saved configuration.

/dm [agent-name]

Opens or creates a Direct Message space with a specific agent. DM spaces are persistent and separate from main chat sessions.

/dm Steve

/channel [name]

Opens or creates a Channel space for multi-agent collaboration.

/channel backend-review

Code intelligence commands

These commands work against your indexed workspace. Run /workspace first to confirm your project is indexed.

/radar

Runs the Proactive Impact Radar against the current git repo. Compares HEAD to its parent, identifies changed files, and reports which other parts of the codebase may be affected.

/radar

Findings are grouped by severity. Requires a git repo at the workspace root with the symbol index populated.

/impact <symbol>

Runs call-graph impact analysis on a named function, type, or identifier. Reports all callers and dependents grouped into HIGH, MEDIUM, and LOW confidence tiers.

/impact validateSession

If no edges are found, Huginn suggests running /workspace to index the repo first.

/workspace

Shows the active workspace state: root path and number of indexed chunks.

/workspace

Run this before /radar or /impact to confirm Huginn has picked up the right project directory.

Session commands

/resume

Opens the session picker overlay. Select a previous session by number or name to restore the full message history.

/save

Force-saves the current session to disk right now. Huginn auto-saves periodically, but use this before closing the terminal if you want certainty.

/title <text>

Renames the current session. Session names appear in the sidebar and in the /resume picker.

/title Auth refactor — March sprint

Sessions are auto-named from your first message but can be renamed at any time.

/notepad

Manages persistent notepads that inject standing context into every agent session.

With no sub-command, the usage summary is shown.

Debug & info commands

/stats

Displays live session statistics: tokens used (input and output), estimated cost, model latency (median and p95), cache hit rate, and files indexed.

/help

Shows the keybindings reference and a summary of slash commands in the current terminal session.

TUI keybindings (also shown by /help):

Navigation commands

These commands navigate directly to a screen with no arguments needed.

You don’t always need a slash

Most things you’d reach for a command to do can be said directly in plain English. Huginn dispatches to the right agent and approach on its own.

have Steve review the auth module
plan a refactor of the database layer
ask Chris to design the new API endpoints
run the test suite and tell me what's failing

Use slash commands when you want explicit control — a specific model, a parallel run, a named agent. Use plain language for everything else.

Skills add commands to the picker

Any skill you install and enable automatically appears in the / picker alongside built-in commands. There is no visual distinction.

huginn skill install <skill-name>

Overriding a built-in: Create a skill with the same name as a built-in. User-installed skills always take precedence over built-ins with the same name — no flags or config required.

See Skills for the full skill system reference.

Full command reference