← Back to blog

v0.4.0: Platform Crons & Session Resume

Agents that schedule themselves, terminal sessions that survive restarts, and the self-sustaining autonomous loop.

v0.4.0: Platform Crons & Session Resume

This release adds the foundation for semi-autonomous agent operation — agents that can schedule their own work, sessions that survive container restarts, and event-driven chaining.

Platform Crons

Platform crons are scheduled prompts sent to local agent terminal sessions on a cron schedule. The agent receives the prompt, acts on it, and the platform tracks execution history.

Key capabilities:

  • Self-scheduling — an agent can call POST /api/platform-crons to create its own recurring tasks
  • Auto-resume — if the target session is dead, the platform re-creates the tmux session and re-launches the agent
  • Execution logging — every cron execution is tracked with timestamps and result status
  • Manual trigger — fire any cron immediately from the UI
# Example: agent creates a health-check cron for itself
curl -X POST http://localhost:8000/api/platform-crons \
  -H "Content-Type: application/json" \
  -d '{
    "name": "health-check",
    "schedule": "*/30 * * * *",
    "prompt": "Check all services are healthy. Report any failures.",
    "agent_name": "claude"
  }'

Session Resume

Terminal sessions now survive container restarts:

  • Orphan cleanup — on startup, active sessions with dead tmux processes are marked closed
  • Resume button — one click to re-create the tmux session and re-launch the agent
  • Last command tracking — the platform remembers what command launched the agent
  • Inline rename — rename sessions directly in the sidebar

Event-Driven Chaining

The automation rules engine now supports send_to_session — when a run completes or fails, the platform can send a prompt directly to an active agent session:

{
  "event_source": "run_event",
  "event_filter": {"event_type": "run_failed"},
  "action_type": "send_to_session",
  "action_config": {
    "tmux_name": "lt-claude-abc123",
    "message": "Run {run_id} on {project_id} failed. Investigate and fix."
  }
}

This enables chained workflows where a failure automatically triggers investigation.

CLI Session Rename

Workspace server CLI sessions now have a rename endpoint (POST /sessions/{id}/rename), matching the local terminal session functionality.

What's Next

  • Platform cron templates (common patterns pre-configured)
  • Cron-to-run bridge (cron fires → creates a full pipeline run)
  • Session sharing between crons and automation rules