Termalin
Safe SSH for you and your AI agents
5 followers
Safe SSH for you and your AI agents
5 followers
A fast cross-platform SSH client (tabs, SFTP with a built-in editor, a database client, session recording) with a built-in MCP server: Claude Code, ChatGPT or any agent works on your servers and never holds your keys, and you decide what it may run. New: server monitoring with no agent to install, and default alerts (disk, memory, services, TLS certificates) to email, push, Telegram or Slack. Free tier, no host limits.
Launch-day update 🚀 — since I first posted this, Termalin grew a built-in database client too. Same custody idea, applied to data: browse and query 13 engines (Postgres, MySQL, Mongo, Redis, ClickHouse, Neo4j…) with a real SQL editor — and when you point an agent at it, DB access is read-only by default and off until you flip it, per connection. The app holds the credential; the agent never gets the connection string. So Claude can answer “why did signups drop Tuesday” by reading the rows, and can't drop a table if it tried. Would genuinely love feedback on the security model — AMA.
Termalin 0.1.45 is out: background jobs and one command on several servers for Claude and ChatGPT; a note for AI on each Connector (what the server is and what to be careful with), which AI apps read before working there; local AI agents can pause while the app is locked (optional); MobaXterm import in the web cabinet, folders included; and a changed server key is no longer a dead end, the web terminal says so and lets you trust the new key. Full notes: termal.in/changelog
Termalin 0.1.47 is out, and with 0.1.46 the app now has an AI assistant: it summarizes a recorded session (what it did, what failed, what's left), explains selected terminal output, and turns a request into a command typed at the cursor (Ctrl+Shift+K) that it never runs for you. In 0.1.47 it can run on your own model: any OpenAI-compatible endpoint, or a local one through Ollama or LM Studio, with your text going straight from your device to that endpoint. Full notes: termal.in/changelog
New in Termalin: server monitoring and alerts, with nothing to install. Every server with a Termalin Connector now gets a dashboard (CPU, memory, disks, network, services, containers, TLS certificates) and alerts with default thresholds: a disk about to fill, a failed service, a certificate about to expire, a server gone offline. They go to email, iOS push, Telegram or a Slack/Discord webhook, once when a problem opens and once when it clears. Claude and ChatGPT can read the same numbers through the hosted MCP endpoint. Part of Pro and Team (and the 14-day trial): termal.in/server-monitoring
Made a 60-second tour of where Termalin is now: the terminal, the toolkit, the MCP server with the key custodian in the middle, the no-open-port Connector and encrypted sync. Sound on: https://youtu.be/1eBDrTDsrIw
Question for you: is a motion-graphics tour like this actually useful, or would you rather see a plain screen recording of the real app?
Termalin 0.1.43 is out: sign in with Google, GitHub or Apple, in the app and on the web; Intel Macs start again (the .dmg is properly Universal now); "Remember password" actually remembers; full screen (F11, Ctrl+Cmd+F on macOS); AI apps now get every command's exit code, so a failed command no longer reads as a success; plus hardening: the Windows key-agent pipe is private to your user, and a locked app refuses local AI tools. Full notes: termal.in/changelog