Termcp: AI-native terminal platform lets agents drive real interactive CLIs without per-tool MCPs
Psychological-Bid722 · reddit · 2026-09-21
A Reddit user released Termcp, an open-source, AI-native cross-platform terminal management platform with a built-in MCP server.
The problem: when agents operate many machines there's no unified access platform; most agents run commands statelessly and non-interactively, so they stall on interactive CLIs (impacket-psexec, msfconsole, etc.); writing a dedicated MCP per interactive tool is a configuration nightmare.
The approach: expose real interactive terminals to agents in a platformized form, so agents can continuously operate a running terminal program instead of just firing off one-off commands—no per-tool MCP development needed.
Human-in-the-loop: visualized session management shows which sessions exist, what each is running, what's on screen, and what the agent last typed. Humans can jump into a session to take over at captchas or dangerous operations, then hand control back to the agent. Installs via go install, aimed at ops, DevOps, and security researchers.
More from coding & agent
- Watching OpenClaw Build OpenClaw: an AI Agent Developing Its Own Harness — vincent_koc · 2026-09-21
- Box CEO: 90% of AI Tokens Will Go to Work No Employee Started Within Five Years — victor_explore · 2026-09-21
- Codex agent confesses its own failure: hid tools, then built machinery to undo it — altryne · 2026-09-21
- Matt Pocock asks which pre-AI codebase design tricks still help agents — mattpocockuk · 2026-09-21
- Dev removes agent tool router after it breaks PR linking and wastes tokens — altryne · 2026-09-21
- FrogNano trains a 4B coding agent to 61.5% SWE-bench via online task synthesis — rohanpaul_ai · 2026-09-21