Skip to content

Play Interface

The Play UI is organized around three regions:

RegionPurpose
Left rosterAgent list, colors, names, status, and quick identity
Center cityAnimated pixel-art environment with agents and the Tool Hub
Right activity panelTimeline, thoughts, tools, communications, and raw logs

Controls

ControlBehavior
STARTLaunches the project run
STOPTerminates the active run without closing the UI
RESTARTStarts a fresh run after a run has started

STOP terminates the child process started by lantern play. On Unix-like systems AgentLantern starts the child in its own process group so related subprocesses can be stopped together.

Agents

Each agent is displayed with:

  • a pixel-art avatar,
  • the agent name below the avatar,
  • a persistent thought bubble above the avatar,
  • an activity badge for task, thinking, tool, delegation, completion, or error state.

When an agent is thinking, the bubble displays ... until the next thought or task event arrives.

Activity Panels

PanelContent
TimelineChronological event stream
ThoughtsCaptured agent thoughts and reasoning snippets
ToolsExternal tool calls, inputs, and associated agents
CommsDelegation and coworker questions
LogRaw stdout and runner messages

Clicking an agent opens its detail panel with event history and completed work.

Final Report

If the run finishes successfully and report.md exists in the project root, AgentLantern reads it and displays it in the UI.

This is common in CrewAI projects with:

yaml
output_file: report.md

Released under the MIT License.