Sage Infrastructure Solutions Group IncSage Infrastructure Solutions Group IncLoading…

Sage Infrastructure Solutions Group Inc

Connect to a host

There are several ways in, and every one of them goes through your access rules and is recorded. (The one exception is break-glass with the host's own credentials, which goes around Keyreeve on purpose — see Break-glass access.)

The desktop app

Most people work in the Keyreeve desktop app, for macOS, Windows and Linux. Download it from the Desktop app page once you are signed in; the page also has an Add to Keyreeve desktop button that fills in your server for you.

  • Tabs on every host you can reach, through the Keyreeve gateway, with keyboard shortcuts for switching between them
  • Kubernetes in the same window: the clusters, namespaces and pods your rules allow, pod logs, recorded pod shells and the kubectl shell
  • A files panel beside the terminal
  • Port forwards to a host or a pod, where your administrators allow port forwarding (it is off by default)
  • Several accounts and servers side by side; with several Keyreeve servers, tabs fail over between them
  • Login scripts, and your own SSH connections to machines outside Keyreeve in a vault sealed with a master password, which also protects the app's own sign-in key

The CLI and your own ssh

Install the keyreeve CLI and sign in. The command to use — with your server's agent service address and CA pin — is on your Home page and the SSH access page; it looks like this:

keyreeve login --server https://agents.keyreeve.example.com:4388 --ca-pin sha256:…

The browser opens, you sign in and approve, and the certificate is loaded into ssh-agent. ssh user@host then works for as long as the certificate lasts (8 hours by default), on hosts that accept direct SSH. The server is remembered, so next time keyreeve login is enough.

Useful flags:

FlagWhat it does
--serverThe Keyreeve agent service URL (otherwise $KEYREEVE_SERVER or the last one used). Works with every command
--ca-pinThe fingerprint of the CA that signed the agent service's certificate, when the server uses a Keyreeve-generated one
--no-browserPrint the link instead of opening a browser
--portFix the local return port, to use with ssh -L
--no-agentSave the certificate to ~/.keyreeve without loading it into ssh-agent
--pasteSign in on another device and paste the code back, if your administrator allows it

keyreeve status shows the certificate you hold; keyreeve logout discards it. For Kubernetes, keyreeve kube login CLUSTER sets up kubectl — see Kubernetes access.

Through the gateway

Some hosts only accept sessions through the Keyreeve gateway, which terminates the SSH connection, relays the session and keeps a server-side recording.

keyreeve ssh-config 'web-*' db01 >> ~/.ssh/config

That writes an ssh config block whose ProxyCommand runs keyreeve proxy for the host, uses your Keyreeve certificate, and trusts Keyreeve's host CA through ~/.keyreeve/known_hosts, so you stop being asked to accept unknown host fingerprints.

The browser

Hosts in the web interface opens a terminal on any host you can reach, relayed through the agent's outbound tunnel. It is covered by your existing sign-in, with no separate step-up.

The Files tab is SFTP over the same login. When your administrators turn on upload scanning, uploads stream through ClamAV into a temporary file that is only put in place once it is clean. Every transfer's path, size, hashes and verdict go into the session's recording and the audit log.

[!NOTE] Terminal history (scrollback) is capped fleet-wide — 1000 lines by default, under Admin → Fleet settings → Terminal history — in the browser and the desktop app alike, and dropped when the session ends (in the desktop app, also when the vault locks). It is the one copy of session output that is not encrypted and access-controlled, so it is deliberately short-lived.

AI assistance

An AI assistant such as Claude can work in a terminal you have opened — run a command and read its output — once you switch on Let AI use this terminal in it. It acts under your own account and permissions, and every AI keystroke is marked in the recording and in the audit log.

Administrators decide where that is possible: AI is off by default, turned on under Admin → Fleet settings (Let people allow an AI to type into their web terminals) or for particular host groups, hosts or Kubernetes clusters, and each access rule says whether the people it covers may use it. Your personal AI tokens for MCP clients are on the AI access page; the desktop app also offers a local MCP server for its own terminals.