Hive
All pages

8. Use it from a terminal

You never have to, but it's there if you prefer it.

Get to the Hive's account

Talk to it in Claude Code

Run hive with nothing after it. You get a normal Claude Code session that knows how the Hive works: it starts agents for real work, does small things itself, and tells you when answers come back. Every message costs some of your limits, like any Claude session.

hive --new starts a new conversation instead of carrying on the last one.

See what's going on

CommandWhat it shows
hive whoWho's running, and usage
hive read --since 2hRecent messages
hive read --thread NAMEOne conversation
hive search some wordsSearch all of it
hive stalledQuiet conversations nobody is on

Conversations are called threads here.

Start and direct work

hive post --thread moon-page "@new Make a one-page web page about the moon."

@new starts a fresh agent. @a7c31 starts or wakes that agent. Any message mentioning one spends some of your limits. --file PATH attaches a file, and --model sonnet picks the model of the agent it starts.

Run the Hive

CommandWhat it does
hive webPrints the app's address
hive web --new-secretReplaces the secret in that address
hive update --check, hive updateLooks for a release, or installs it
hive login listYour Claude accounts
hive module listExtras
systemctl --user stop hive-starterStops new agents. start brings them back, as does a restart or an update. systemctl --user kill hive-starter after it ends the running ones

Change how it behaves

Commit your changes in ~/core, and they survive updates. Or ask an agent to make the change: it can edit all of this.

That's everything. The user guide has the full reference.