Since January, I've used AgentTerm as my only terminal for all my projects, including AgentTerm itself.
AgentTerm expands each AI CLI session into its own rich visual workspace, keeping the familiar terminal interface and existing CLI functionality. As AI takes on more implementation, I focus more on plans and reviews.
When an agent writes a plan, I open it in the terminal window and write down my intent. The agent replies and edits the plan in place.
Before every PR at work, I use a curated review. The agent presents the changes that need my judgment, explains its reasoning, and includes relevant unchanged code for context. I comment on the code or its explanation, and it fixes and replies.
I also use it to recognize and resume sessions and continue from my phone. These are examples of broader improvements that work together. The visual parts appear when needed and get out of the way afterward.
You and your agents can extend it further to fit your specific needs or those of your team or organization.
MIT licensed, no telemetry or account. macOS and Windows through WSL. The repo has the demo and setup instructions, including separate setup for phone access and IDE integration.
Comments URL: [https://news.ycombinator.com/item?id=49929891](https://news.ycombinator.com/item?id=49929891)
Points: 1