Conduit vs uniTerm

Conduit is a remote connection manager that brings SSH terminals, RDP desktops, VNC sessions, and web tools into a tabbed workspace. It includes a native MCP server with tools for terminal, RDP, VNC, web, credential, and vault operations. Conduit can launch Claude Code, Codex, Grok Build, Cursor Agent, OpenClaw, Gemini CLI, GitHub Copilot, and OpenCode; other MCP-compatible clients can connect. AI inference is not paid for or proxied by Conduit, so an agent runs under the user's provider subscription.

Connection entries and credentials are stored in a local SQLite vault encrypted with AES-256. Encryption keys are derived locally from the master password and are not transmitted to Conduit servers, including for cloud backups. MCP calls use local sockets or named pipes, redact secrets, apply per-tool rate limits, log activity, and require one-off approval for credential reads. Free accounts and local-mode installs are capped at 50 MCP tool calls per day. The free plan includes unlimited connections; Pro is free for 30 days, then costs $10 a month. Desktop versions are available for macOS, Windows, and Linux, with iPhone and iPad companion apps.

uniTerm is a free, open-source terminal emulator for Windows, macOS, and Linux, with a built-in AI Agent that can plan and execute multi-round shell commands. It supports connections and tools across protocols including SSH, Telnet, Mosh, serial, file transfer, remote desktops, databases, containers, and server monitoring. Its AI sidebar works with Anthropic and OpenAI-compatible APIs, including Claude and GPT. Command execution can be set to run without confirmation or to require approval for every command. Connection management can import setups from Xshell, MobaXterm, DBeaver, Navicat, and virt-viewer. The guide describes encrypted synchronization of connections, favorites, and AI settings through a private GitHub, GitLab, or Gitee repository. A credential keystore stores reusable login identities, including passwords and private keys. Its MCP server lets external AI agents run commands and transfer files on saved connections; the project README says credentials never leave the app. The server monitor supports only remote Linux servers. Telnet sends data without encryption and is cautioned for trusted networks. The project is licensed under Apache 2.0 and described as a personal hobby project maintained in the author's spare time.