Skip to content

Tour

A walk through the terminal UI. Every image is from a recorded run; click to enlarge.

Hub

agent6 tui lists every run for the repository with its mode, status, last activity, and cost. Enter opens a run; c opens the config page; ? lists the keys.

The hub

Conversation

Opening a run lands on its conversation (also agent6 sessions transcript): the task, the model's reasoning, and every tool call with its complete input and output, following live. A live run keeps a steer bar at the bottom.

A run transcript

Run dashboard

Ctrl+D toggles the dashboard: the task graph beside the model's live reasoning, then the tool calls with their results, and the event log and latest commit diff side by side. The composer bar runs along the foot (type to steer, or to resume a finished run with a follow-up; / completes the steer directives, Ctrl-R searches the session's past messages), and the View menu maximizes the focused pane to full screen.

The run dashboard

Event log

The View menu's Full log opens the JSONL event stream the dashboard is built from, scrollable over the whole run.

The event log

Configuration

The config page shows every setting, its effective value, and where that value came from (a built-in default, the global config, or the per-repo config). / filters by name.

The config page

Filtering the config by name

Keys

The keys and actions overlay

Watching a state machine

Beyond one-shot runs, agent6 runs editable state machines: a .asm.toml of tool, branch, agent, and wait states driven over a journal. Author one, read its graph, and watch it execute. Here code-fixer runs a fix-loop: an agent state edits the repo to make a failing check pass, a tool state re-runs the check, and the machine routes on the result until it is green or the attempt budget is spent, with the agent's reasoning streamed live like a run.

The same machine from the browser: start it on the Machines page and watch the current state stream, with the run's approvals and questions answerable in place.

From the terminal

For terminal-first workflows, agent6 run executes in the foreground: steer it with Ctrl-C, no TUI required (the pause menu Tab-completes its commands; Up recalls and Ctrl-R searches the session's past messages). Runs started from the TUI or web hub are detached instead, and agent6 attach attaches to either kind: the run's conversation by default, --raw for a no-deps line tail, --tui for the full-screen TUI, --json for a one-shot snapshot of the same state.

These are regenerated from recorded runs by the pages workflow, so they track the current UI.