Threads

A thread is one conversation with one agent, tied to one working directory. It's the basic unit you interact with in LinkCode — every message, tool call, and permission prompt you see belongs to exactly one thread.

Status

A thread moves through a small set of states as you use it:

• Starting — the agent process is being launched.
• Idle — the agent is ready and waiting for you to say something.
• Running — the agent is actively working on your last message.
• Awaiting input — the agent has stopped to ask you something, most often a permission request.
• Stopped — there's no live agent process behind this thread right now. The thread and its history are still there; nothing is currently running.

Creating a thread

Start a new thread from the New Task button at the top of the sidebar, or open the command palette (⌘K on macOS, Ctrl+K elsewhere) and use New thread. Either way you pick an agent and a working directory — see workspaces for how directories are registered.

If the directory is a git repository, you can also pick a branch. Choosing one starts the thread in an isolated git worktree for that branch, so the agent's work doesn't disturb whatever is checked out in your main copy.

Stopping vs. closing

These sound similar but do different things:

• Stopping a thread ends the agent's live process but keeps the thread in your sidebar, now in the Stopped state. You can resume it later.
• Closing a thread removes it from LinkCode's list entirely (and stops any live run first). This does not touch the agent's own transcript on disk — the underlying conversation history the agent itself keeps is untouched, so a closed thread can still be brought back later through import.

Pin keeps a thread at the top of the sidebar. The thread's More actions menu also copies its title, reveals its working directory in your file manager, and opens that directory in your editor.

Import history

Agents keep their own conversation history outside of LinkCode — everything you've run through the claude CLI, for example. Import chat history browses those existing conversations and pulls them into LinkCode as threads, one at a time or all at once, so you can keep working on them from the LinkCode interface.

Claude Code, Codex, OpenCode, and Pi can all list, read, and resume their own history this way. Grok Build cannot: its threads start fresh. See the agents overview for the full comparison.