One Terminal to Rule Them All: uniTerm Puts SSH, RDP, Kubernetes, and an AI Agent in the Same Window

If you're anything like me, your daily workflow involves a small army of terminal windows: one for SSH, another for a database client, a third for Kubernetes, plus a browser tab for that RDP session you keep forgetting to close. What if you could consolidate all of that into a single app—and have an AI agent handle the tedious command sequences for you? That's the pitch behind uniTerm, an open-source all-in-one terminal that supports over 30 protocols and ships with a built-in autonomous AI assistant.

What It Does

uniTerm is a lightweight terminal application that covers remote access needs across the board. On the remote terminal side, it supports SSH, Telnet, Mosh, and Raw TCP, with password or key authentication, plus SSH tunnel port forwarding so any connection can route through a jump host. For local work, you get PowerShell, CMD, Git Bash, and WSL, along with serial connections that let you configure baud rate, data bits, stop bits, parity, and local echo.

Beyond basic shell access, uniTerm bundles in file transfer (SFTP, SCP, FTP, FTPS, SMB, WebDAV, S3, and Zmodem, with dual-pane browsing and rz/sz support), remote desktop (RDP, VNC, SPICE, and X11), database clients (MySQL, PostgreSQL, Oracle, SQL Server, rqlite, Redis, MongoDB, Elasticsearch), container management (Kubernetes, Docker, Podman, nerdctl, WSLC), and a server monitor showing real-time CPU, memory, disk, network, processes, ports, and interfaces.

The AI side is where things get interesting. uniTerm includes an autonomous agent that plans and executes multi-turn shell commands directly in your terminal. It integrates with Anthropic and OpenAI-compatible APIs (so Claude, GPT, and other compliant models work), saves chat history per session, and offers four execution modes—bypass, dangerous only, dangerous plus write, or confirm all—so you decide how much oversight the agent needs. Commands run in the active terminal tab, or you can pin the agent to a specific tab. Split panes let you collaborate side-by-side, each with its own terminal context. There's also smart completion for SSH terminals while you type.

The app runs on Windows, macOS, and Linux, and it's licensed under Apache 2.0.

Why It's Cool

• The AI agent actually runs commands, not just suggests them. This isn't a chatbot that spits out a kubectl incantation you copy-paste yourself. The agent plans, executes, observes results, and iterates across multiple rounds without manual intervention. That's a meaningfully different interaction model—closer to delegating than to asking for help.

• You control the blast radius. Handing an AI agent shell access sounds terrifying, and it should. uniTerm's four execution modes let you dial in exactly how much autonomy the agent gets. "Confirm all" keeps you in the loop on every command. "Bypass" gets out of the way entirely. The dangerous-only and dangerous-plus-write modes sit in between. That's a thoughtful design choice that acknowledges the real risk here.

• Thirty-plus protocols in one binary is genuinely useful. Most people don't need all of them, but the ones you do need are probably scattered across four or five separate applications right now. Consolidating SSH, RDP, database clients, and Kubernetes tooling into one app reduces context-switching and, more practically, reduces the number of apps you need to keep updated.

• Split panes with independent terminal contexts. Each pane has its own terminal context, which means you can run the AI agent in one pane against one server while working manually in another. That's a small detail that makes a big difference in practice.

• Persistent conversations per session. Chat history survives app restarts. If you're mid-investigation on a production issue and need to close your laptop, you don't lose the thread.

How to Try It

1. Head to the GitHub repository or the official homepage.
2. Grab the latest release from the releases page—builds are available for Windows, macOS, and Linux.
3. If you'd rather build from source, the README includes a "Build from Source" section with instructions.
4. Once installed, configure your LLM integration (Anthropic or any OpenAI-compatible API) to enable the AI agent.
5. Set your execution mode. If you're trying the agent for the first time, "confirm all" is the sensible starting point.

There's also a user guide if you want to dig into the details.

Final Thoughts

uniTerm is aiming at a specific kind of user: someone who juggles a lot of remote infrastructure and wants fewer windows open. The protocol coverage is broad enough that most sysadmins, DevOps engineers, and backend developers will find their daily tools represented. The AI agent is the headline feature, and the execution modes suggest the developers have thought carefully about how to make it useful without making it reckless. It won't replace your specialized tools for every edge case, but as a daily driver for common remote work, it's worth a look. The project is Apache 2.0 licensed, actively developed, and has a roadmap in the README if you want to see where it's headed.

Follow @githubprojects for more developer tools and open source projects.