hydra-cli
hydra-cli is for a computer where you do not keep a browser open: an office PC, a server, a machine on a network that only it can reach. Pages are read from that computer's network. It has no browser and no logins, so sites that need your sign-in are a job for Hydra for Chrome and Firefox.
Install
Download the file for your system from Devices → Get the companion. In the commands below, replace
https://hydra.example.com with your company's Hydra address.
macOS
chmod +x hydra-cli-*-macos-*
xattr -d com.apple.quarantine hydra-cli-*-macos-* # allow a file downloaded outside the App Store
./hydra-cli-*-macos-* login --url https://hydra.example.com
Windows (PowerShell)
.\hydra-cli-...-windows-x64.exe login --url https://hydra.example.com
Linux (x64, or ARM64 such as a Raspberry Pi)
chmod +x hydra-cli-*-linux-*
./hydra-cli-*-linux-* login --url https://hydra.example.com --no-browser # --no-browser over SSH
login prints a code and opens Hydra's approval page. Approve it there and the computer is paired. On a machine without a screen,
add --no-browser and open the printed link on any device where you are signed in to Hydra.
Check which account it is paired to
login ends with Paired to name <email>, and hydra-cli status prints the same on its account: line. If it shows someone
else, run hydra-cli logout (or Revoke the device in Devices) and pair again.
Keep it running
hydra-cli service install # start at login (macOS LaunchAgent, Linux systemd user unit, Windows task at logon)
hydra-cli status # paired? reachable?
hydra-cli run # run in the foreground instead
Where the token is kept
The pairing token goes to the system keychain (macOS Keychain, Windows Credential Manager, Linux Secret Service). On a server
without one, set HYDRA_CLI_TOKEN_STORE=file: the token is kept in a file only you can read. Use the same setting for login,
run and logout.
What it reads, and what it refuses
- Pages your chats ask for, read with a plain GET from this computer, at most 3 MB each.
- It refuses private and local addresses (routers, NAS,
localhost) on every redirect, so a page cannot steer it into your LAN. - It never acts on a page: no forms, no posts.
Move your Claude and ChatGPT history
hydra-cli migrate copies your chats from this computer into your own Hydra account. You choose what moves, and nothing is
sent until you agree.
hydra-cli migrate scan # show what this computer holds; sends nothing
hydra-cli migrate start # the choice screen, then the move
hydra-cli migrate start --chatgpt-export chatgpt.zip --claude-export claude.zip
hydra-cli migrate start --web # agree in the browser instead of the terminal
hydra-cli migrate status --follow # one line per second
hydra-cli migrate pause # also: resume, stop
You agree first. migrate start shows one line per source with a tick box: Claude Cowork chats on this Mac, Claude Code chats,
a ChatGPT export file and a claude.ai export file. Type a number to tick or untick, press Enter to start, or q to cancel.
Pause, resume, stop. Ctrl-C pauses: the chunk in flight finishes, a checkpoint is saved, and hydra-cli migrate resume
continues from it. Stop deletes the staging files on this computer and what Hydra holds; chats already imported stay as ordinary
chats. One migration runs at a time.
What is never taken. Passwords, sign-in tokens and browser data. Keys and passwords found in text (API keys, access tokens,
private-key blocks) are replaced by [redacted] on this computer, and the count is shown. Data is packed and encrypted on this
computer and sent only to your own account.
Where to get the exports. ChatGPT: Settings → Data controls → Export. claude.ai: Settings → Privacy → Export. In Hydra for Mac and Windows, the menu item Move my Claude & ChatGPT to Hydra... opens the same move.
Remove it
hydra-cli logout # revokes it in Hydra and deletes the token here
Or Revoke it in Devices.