# Critters: An Automated Issue Processing Daemon Based on Claude Code

> Introducing a TypeScript daemon that monitors tagged Issues in Linear or Jira and automatically runs Claude Code workflows, enabling end-to-end automation from planning to execution and review.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-04-19T05:15:14.000Z
- 最近活动: 2026-04-19T05:19:45.615Z
- 热度: 159.9
- 关键词: Claude Code, 自动化工作流, Issue跟踪, AI代理, TypeScript, Linear, Jira, DevOps
- 页面链接: https://www.zingnex.cn/en/forum/thread/critters-claude-codeissue
- Canonical: https://www.zingnex.cn/forum/thread/critters-claude-codeissue
- Markdown 来源: floors_fallback

---

## [Main Post/Introduction] Critters: An Automated Issue Processing Daemon Based on Claude Code

Introducing Critters—a TypeScript daemon that monitors Issues with specific tags in Linear or Jira, automatically runs Claude Code workflows, and enables end-to-end automation from planning and execution to review. It integrates AI programming tools with issue tracking systems, supports custom workflows and multiple deployment methods, focuses on operation and maintenance experience and security management, and represents the direction of end-to-end automation in AI-assisted software development.

## Industry Background: Efficiency Bottlenecks in Development Workflows and the Potential of AI Agents

The software engineering field has long faced efficiency bottlenecks such as issue backlogs and mismatched development resources. Stripe's open-source Minions project demonstrates the potential of AI agents to automatically handle development tasks. As an open-source implementation of this concept, Critters builds an automated workflow for autonomous planning, execution, and code review by integrating AI tools like Claude Code with issue tracking systems.

## System Architecture and Design Philosophy: Daemon and Observability

Critters adopts a daemon architecture, built on the Bun runtime, and manages multiple concurrent tasks via tmux panels. Bun ensures task processing efficiency, while tmux supports real-time monitoring of AI agent execution. The system supports Linear and Jira, detecting Issues with specific tags through polling and triggering predefined workflows.

## Workflow Stages: Configurable Task Decomposition Mechanism

The core abstraction is the Critter type, where each type defines configurable workflow stages (default includes planning, execution, review, and supports customization). Each stage corresponds to a Claude Code session, automatically passing context such as issue descriptions, repository status, and branches. The phased design decomposes complex tasks while retaining the flexibility for human intervention.

## Command-Line Tools: Full-Lifecycle Operation and Maintenance Support

It provides a rich set of command-line tools covering initialization to daily operation and maintenance: the start command creates a tmux session to enter daemon mode; the status command displays an overview of active/queued tasks; the log command supports viewing details by stage or real-time stream; the prompt-render command previews prompt templates to help debug and optimize AI behavior, reducing deployment barriers.

## Deployment Options: Flexible Adaptation to Different Infrastructures

Multiple deployment options are available: binary installation is suitable for quick trials; Docker Compose for containerized deployment with built-in dependencies like Claude Code CLI and GitHub CLI; supports custom Dockerfiles to extend language runtimes (e.g., Node.js, Bun, Python).

## Security Policies: Sensitive Information and Permission Control

API keys (Anthropic, Linear/Jira authentication information) are injected via environment variables; Slack notifications support threaded pushes via Webhook URLs or Bot Tokens; documentation reminds users to manage sensitive configurations using .env files to avoid hardcoding keys into code or configurations.

## Future Outlook: Evolution Direction of AI-Assisted Development

Critters represents the direction of AI-assisted development moving from code completion to end-to-end task automation. AI agents take on repetitive, context-clear tasks, while humans focus on architectural design, complex decision-making, and innovation. With the advancement of multimodal large models, such automated workflows are expected to become the new normal in software engineering.