FAQ

How is this different from running Claude Code in a terminal?

You still run Claude Code in a real terminal — spekterm adds what a row of terminal tabs does not have. Each repository gets its own place on the rail with its sessions, and next to the terminal is a side panel that understands OpenSpec: it shows the change the agent is working on — its proposal, design, tasks, and specs — and follows along as the agent writes to disk. Sessions also survive a restart, and an agent can hand work to another repository. See The OpenSpec side panel.

How is it different from tools that run many agents in parallel?

Those tools focus on running many agents at once. spekterm focuses on keeping each session next to the OpenSpec change it is working on, so you can read the spec while the agent writes the code. It runs the real claude CLI in a real terminal and never scrapes the screen; the conversation view is built from the agent’s own transcript.

Do I need OpenSpec?

No. Terminals, agent sessions, the conversation view, files, and the inbox work in any folder. The side panel’s OpenSpec view simply has nothing to show in a folder without an openspec/ directory.

Does it work on macOS or Windows?

Not yet — builds are available for Linux today. A macOS build is tracked in issue #63.

Is my code sent anywhere?

spekterm itself does not send your code anywhere. The programs it starts for you — such as claude — connect to their own services. See Data and network for the full list.

Is spekterm affiliated with Anthropic or OpenSpec?

No. spekterm is an independent open-source project. Claude Code is Anthropic’s product and OpenSpec is its own project; spekterm only runs them.

Is it free?

Yes. spekterm is open source under the MIT license. Claude Code runs with your own subscription.