AutoDev is a fully automated AI-driven software development framework designed to enable autonomous planning and execution of complex software engineering tasks. It addresses the limitations of existing AI solutions, which often focus on code snippet suggestions and file manipulation within chat interfaces. AutoDev allows users to define complex software engineering objectives, which are then assigned to AI agents that can perform a wide range of operations on a codebase, including file editing, retrieval, build processes, execution, testing, and git operations. These AI agents have access to various resources such as files, compiler output, build and testing logs, and static analysis tools, enabling them to execute tasks with a comprehensive understanding of contextual information. AutoDev ensures a secure development environment by confining all operations within Docker containers and incorporating guardrails to protect user privacy and file security. Users can define specific permitted or restricted commands and operations within AutoDev. The framework includes several key features: 1. **Conversation Manager**: Manages user and AI agent conversations. 2. **Tools Library**: Provides a variety of code and SE-related tools. 3. **Agent Scheduler**: Orchestrates AI agents to collaborate on tasks. 4. **Evaluation Environment**: Executes code and runs tests within a secure Docker environment. In their evaluation, the authors tested AutoDev on the HumanEval dataset, achieving promising results with 91.5% and 87.8% Pass@1 for code generation and test generation, respectively. These results demonstrate AutoDev's effectiveness in automating software engineering tasks while maintaining a secure and user-controlled development environment. The paper also discusses the design of AutoDev, including the configuration of rules and actions, the role of the Conversation Manager, the Agent Scheduler, and the Tools Library. The evaluation section details the effectiveness of AutoDev in code and test generation tasks, comparing it to other approaches and highlighting its efficiency in completing tasks. Finally, the authors discuss future plans, including integrating AutoDev into IDEs and CI/CD pipelines, and exploring multi-agent collaboration and human feedback mechanisms.AutoDev is a fully automated AI-driven software development framework designed to enable autonomous planning and execution of complex software engineering tasks. It addresses the limitations of existing AI solutions, which often focus on code snippet suggestions and file manipulation within chat interfaces. AutoDev allows users to define complex software engineering objectives, which are then assigned to AI agents that can perform a wide range of operations on a codebase, including file editing, retrieval, build processes, execution, testing, and git operations. These AI agents have access to various resources such as files, compiler output, build and testing logs, and static analysis tools, enabling them to execute tasks with a comprehensive understanding of contextual information. AutoDev ensures a secure development environment by confining all operations within Docker containers and incorporating guardrails to protect user privacy and file security. Users can define specific permitted or restricted commands and operations within AutoDev. The framework includes several key features: 1. **Conversation Manager**: Manages user and AI agent conversations. 2. **Tools Library**: Provides a variety of code and SE-related tools. 3. **Agent Scheduler**: Orchestrates AI agents to collaborate on tasks. 4. **Evaluation Environment**: Executes code and runs tests within a secure Docker environment. In their evaluation, the authors tested AutoDev on the HumanEval dataset, achieving promising results with 91.5% and 87.8% Pass@1 for code generation and test generation, respectively. These results demonstrate AutoDev's effectiveness in automating software engineering tasks while maintaining a secure and user-controlled development environment. The paper also discusses the design of AutoDev, including the configuration of rules and actions, the role of the Conversation Manager, the Agent Scheduler, and the Tools Library. The evaluation section details the effectiveness of AutoDev in code and test generation tasks, comparing it to other approaches and highlighting its efficiency in completing tasks. Finally, the authors discuss future plans, including integrating AutoDev into IDEs and CI/CD pipelines, and exploring multi-agent collaboration and human feedback mechanisms.