What problem does it solve?
Provides a structured harness to turn Claude Code into a persistent, self-directing agent system that can run on schedules, maintain cross-session memory, and perform automated workflows without manual intervention.
Core Features & Use Cases
- Persistent Memory: Combine built-in project memory files with an MCP memory server for short-, medium-, and long-term state and structured knowledge.
- Scheduled Operations (Crons): Define recurring tasks to run agent sessions on a schedule for monitoring, reporting, or maintenance.
- Dispatch and Remote Agents: Trigger agents from webhooks, CI/CD, or other agents for event-driven automation.
- Computer Use Integration: Optionally automate browser and desktop interactions for end-to-end workflows like UI testing or form entry, with explicit permission gating.
- Task Queue & Orchestration: Maintain a durable task queue persisted to project memory to survive session boundaries and coordinate multi-step pipelines.
- Example Workflows: Autonomous PR reviewer, daily research digest, and pre-meeting briefing agents.
Quick Start
Create a scheduled task that reviews open PRs for a watched repo every weekday morning and writes a summary to your project memory.
