Enough to get productive. Anything it doesn't answer, ask — that also tells us what belongs here.
Click New (or press ⌘N), pick a connection type, and fill in the server details. Nineteen types are built in: nine databases (MySQL, PostgreSQL, SQL Server, Oracle, SQLite, ClickHouse, Athena, DuckDB, DynamoDB), file and storage services (SFTP, FTP/FTPS, Amazon S3, WebDAV, SMB, Azure Blob Storage, local folders), SSH terminals, Git repositories, and Subversion working copies. Test Connection works for every one of them before you save.
Passwords and keys go to the macOS Keychain, never into the app's database. The first time you reach an SSH or FTPS server the app shows you its host key or certificate and asks; it never trusts a server silently, and it warns loudly if a server's identity later changes.
Connections live in workspaces — group them by client, environment, or however you think. A credential (an "account") can be shared by several connections, so rotating one password is one edit. More on the library page.
Opening a file connection shows two panels: your Mac on the left, the server on the right. Drag either way to copy, folders included. Transfers queue in a drawer with per-file progress and can be cancelled mid-file. A dropped file never overwrites silently — it takes the first free "name 2"-style name instead.
New Folder, Rename and Delete live in the context menu wherever the protocol supports them. Deleting a folder is recursive and stops at the first thing that will not delete, reporting what happened rather than claiming success. To check a server folder against your local copy, right-click it and choose Compare with Local. More on the files and storage page.
Double-click a remote file (or choose Edit) and it opens in a tab: syntax highlighting for 36 languages, line numbers, find and replace, go to line. Saving is protocol-aware — SFTP uploads to a temporary name and atomically renames over the original, S3 sends a conditional write the server enforces, and where a protocol can only plain-overwrite, the app says so instead of pretending. If someone else changed the file while you had it open, the save is refused and you are shown both versions; nothing is silently clobbered. Unsaved edits survive a crash. More on the editor page.
A database connection opens a schema tree — tables, views, stored procedures, functions, and per-table indexes, foreign keys and triggers where the engine has them. Browse table contents, run queries, and export any result set as CSV, tab-separated, JSON, Markdown, or SQL INSERT statements.
A SQL Server connection can sign in with Microsoft Entra ID instead of a SQL login: choose it under Authentication in the connection editor. The browser sign-in handles multi-factor authentication, and Hydra keeps the sign-in on that device so you are asked again only when it expires. More on the databases page.
A Git connection shows working-tree status, history, and diffs; stage or discard by file or by hunk, commit, push and pull. An SSH connection opens a real terminal in a tab, reusing the same host-key trust the file browser established. More on the Git and SSH pages.
Choose File › Compare Files or Folders…, Compare Clipboard with File… or New Text Comparison, or run hydra diff A B in Terminal. Two folders list every file that differs; select one to see its changes side by side or unified. Copy a change across with the arrow beside it (⌥⌘← and ⌥⌘→) and save; the window asks before closing with anything unsaved. Images compare as pictures, other binary files as hex, and any comparison can be copied or saved as a patch. Either side can be on a server: hydra diff web:/var/www ./site.
When Git stops on a conflict, right-click the file in the Git tab and choose Resolve Conflicts…: Yours, the Result and Theirs side by side. Answer each conflict with a click, or double-click it to type the answer yourself. Save is available once every conflict is answered. hydra git-setup --apply makes Hydra the tool git difftool and git mergetool open. More on the compare and merge page.
Point the scanner at a project folder and it finds connection details in config files — fourteen formats. An on-device model judges whether a found secret looks real or is a placeholder; file contents never leave your Mac. Nothing is imported until you select it.
AI coding agents such as Claude Code, and your own scripts, can use your connections through Hydra: list them, browse and transfer files, run queries. They are never handed a credential; the app holds the secrets and does the connecting. It is off until you choose otherwise under External Agent Access (asked on the welcome screens, and in Settings › Automation): Always, While Hydra is Open, or Never.
Each program is approved by you the first time it connects, each connection can be made read-only to agents, and every call, including failures, is recorded in the Automation Log. Settings › Automation › Connect an Agent copies the one line that registers Hydra with Claude Code. The details are in the agents and automation FAQ and on the automation page.
Help ▸ Keyboard Shortcuts lists every binding, and Settings ▸ Shortcuts lets you change them.
Workspace secrets (API tokens and the like) live in the Keychain too; each one has a copy button in the Secrets list and its editor, and a copied value is hidden from clipboard history apps and cleared after 90 seconds. The connection library is on your Mac and, if you use iCloud, in your private iCloud database. Passwords stay in the Keychain. The details are on the privacy page, which is short because there is little to disclose.
What each setting does, which ones follow you to your other Macs, and the options in the connection editor (access, encryption, tunnels, trusted keys) are in the settings FAQ.
Help ▸ Email Support pre-fills your app version so you don't have to find it. Or start at the support page.