# Agentry: A Declarative AI Agent Workflow Orchestration Tool for Seamless Migration Between Local and CI Environments

> Agentry is an open-source CLI tool for AI agent workflow orchestration, enabling developers to define AI workflows using YAML and run them consistently in both local and CI environments. Its core advantages lie in a security-first design philosophy, including sandbox isolation, output validation, and least-privilege execution.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-03-28T07:46:04.000Z
- 最近活动: 2026-03-28T07:49:05.224Z
- 热度: 154.9
- 关键词: AI Agent, Workflow Orchestration, LLM, CI/CD, YAML, Security, Claude Code, GitHub Actions, Automation, DevOps
- 页面链接: https://www.zingnex.cn/en/forum/thread/agentry-ai-ci
- Canonical: https://www.zingnex.cn/forum/thread/agentry-ai-ci
- Markdown 来源: floors_fallback

---

## Agentry: Declarative AI Agent Workflow Orchestration Tool for Seamless Local-CI Migration

Agentry is an open-source CLI tool for AI agent workflow orchestration, allowing developers to define workflows via YAML and run them consistently in local and CI environments. Its core advantages include a security-first design (sandbox isolation, output validation, least privilege execution) and the ability to generate GitHub Actions pipelines from the same workflow definition.

## Pain Points in AI Workflow Orchestration

With LLM capabilities growing, teams build AI agent-based workflows, but often face issues: workflows running well locally need re-implementation in CI/CD, leading to doubled maintenance costs and inconsistent behavior. Agentry addresses this problem.

## Core Components of Agentry's YAML Workflow

Agentry's workflow uses seven modules: 
1. Identity: Name, version, description for version control. 
2. Inputs: Supports git-diff, repository-ref, auto-parsing references like `diff=HEAD~1`. 
3. Tools: Explicit tool manifest for least privilege (agents only access listed tools). 
4. Agent Config: Specifies runtime (Claude Code currently), model version, system prompt path (config-execution separation). 
5. Safety: Timeout, trust level (sandbox via Docker or privileged mode). 
6. Output: JSON Schema validation for reliable downstream consumption. 
7. Composition: DAG pipelines with concurrent nodes, failure strategies (abort/skip/retry).

## Five-Layer Security Architecture of Agentry

Agentry's security has five layers: 
1. Definition layer: YAML parsed via Pydantic for structural validity. 
2. Security layer: SecurityEnvelope enforces trust levels, pre-checks, signature validation. 
3. Parsing layer: EnvironmentBinder maps abstract inputs/tools to local/GitHub Actions implementations. 
4. Execution layer: RunnerProtocol provides isolation (DockerRunner/InProcessRunner). 
5. Agent layer: AgentProtocol delegates to runtime like ClaudeCodeAgent. 
Also, Ed25519 signature for safety/output modules to prevent tampering; `agentry validate --security-audit` audits changes.

## Practical Use Cases of Agentry

Agentry has built-in examples: 
- Code review: Analyze PR diffs for vulnerabilities, performance issues, style violations, comment results on PR. 
- Issue classification: Smartly categorize and route new issues. 
- Bug fix: Diagnose issues, generate fixes, auto-create branches/PRs (needs manual review). 
- Task decomposition: Split complex requirements into executable tasks for project planning.

## Installation and Key Commands for Agentry

Requirements: Python3.10+, Claude Code installed (authenticated), Docker optional (for sandbox). 
Installation: `pip install agentry`. 
Key commands: 
- `agentry validate <workflow>`: Validate definition. 
- `agentry run <workflow>`: Local execution. 
- `agentry ci generate --target github <workflow>`: Generate GitHub Actions config. 
- `agentry keygen`: Generate signature keys. 
- `agentry sign <workflow>`: Sign workflow.

## Extensibility and Conclusion of Agentry

**Extensibility**: Binder system uses Python entry points for plugins (add new CI targets via EnvironmentBinder protocol; new agent runtimes via AgentProtocol). 
**Conclusion**: Agentry elevates AI workflows from ad-hoc scripts to maintainable, auditable, portable infrastructure. It's a practical choice for teams wanting AI integration with flexibility and security, representing an important direction in AI engineering.