Skip to content

Dashboard ​

The dashboard provides a real-time overview of all your AI coding agents.

Agent Cards ​

Each agent is displayed as a rich information card showing:

  • Provider badge: CLAUDE (orange) or CODEX (green)
  • Agent name: Click to enter chat view
  • Status indicator: Running (green), Stopped (gray), Error (red), Waiting Input (yellow)
  • Project & git branch: Which repository and branch the agent is working on
  • Pull Request link: If the agent created a PR, a direct link is shown (auto-detected from agent output)
  • Model & context usage: Which LLM model the session uses and a visual progress bar showing context window consumption
  • Task description: A summary of what the agent is currently doing (from the initial prompt)
  • MCP servers: Which Model Context Protocol servers are connected to the session (parsed from --mcp-config)
  • Cost: Total API cost in USD (Claude) or token count (Codex)
  • Latest message: Preview of the most recent agent response

Actions ​

  • + New Agent: Navigate to the agent creation form
  • Stop All: Stop all running agents at once
  • Delete: Remove a stopped internal agent. If configured as ask or do not purge, a confirm dialog appears and lets you choose whether to purge CLI session files by sessionId
  • Settings: Configure auto-delete retention for stopped internal agents and default session-file delete policy (ask / do not purge / always purge)

External Agent Discovery ​

Agent Monitor automatically discovers Claude Code and Codex processes running outside the dashboard (e.g., started from a terminal). These are displayed with an EXT badge while their local process is alive.

  • Automatic session ingestion: Existing local sessions are loaded from provider log files (~/.claude/projects/**.jsonl and ~/.codex/sessions/**.jsonl) so history appears automatically after discovery.
  • Running-only visibility: External cards are removed automatically after the underlying process exits.
  • Safe deletion model: External cards cannot be deleted from Agent Monitor.
  • Toggle visibility: Click the External (N) button in the dashboard toolbar to show or hide external agents. The preference is persisted across sessions.
  • Live updates: External agents stream messages and status changes in real time, including tool calls/results and token/context updates when available.
  • Source indicator: External agents are marked with source: 'external' and display an "EXT" badge on their card.
  • Internal agents are unaffected: Internal agents created by Agent Monitor remain visible after stop (until manual delete or retention cleanup).

Auto-Delete Expired Agents ​

Stopped internal agents can be automatically cleaned up after a configurable retention period. Open Settings from the dashboard to set the retention time in hours (default: 24 hours). Set to 0 to keep agents forever. The server checks for expired agents every 60 seconds.

Real-time Updates ​

The dashboard uses Socket.IO for live updates. Agent status changes, new messages, and cost updates stream instantly without refreshing the page via agent:update (per-agent room events) and agent:snapshot (broadcast dashboard updates).