AutoDev is an AI-driven software development framework that enables autonomous planning and execution of complex software engineering tasks. It allows users to define software engineering objectives, which are then assigned to autonomous AI agents. These agents can perform a variety of operations on a codebase, including file editing, retrieval, build processes, execution, testing, and git operations. AutoDev provides a secure development environment by confining all operations within Docker containers and includes guardrails to ensure user privacy and file security. The framework also allows users to define specific permitted or restricted commands and operations within AutoDev. AutoDev was evaluated on the HumanEval dataset, achieving a Pass@1 score of 91.5% for code generation and 87.8% for test generation, demonstrating its effectiveness in automating software engineering tasks while maintaining a secure and user-controlled development environment. The framework includes a Conversation Manager, a Tools library, an Agent Scheduler, and an Evaluation Environment. The Conversation Manager tracks and manages user and AI agent conversations, while the Tools Library provides a range of commands that empower agents to perform diverse operations on the repository. The Evaluation Environment allows secure execution of file editing, retrieval, build, execution, and testing commands. AutoDev's design enables autonomous AI agents to perform complex software engineering tasks without requiring developer intervention beyond setting the initial objective. The framework supports multi-agent collaboration, allowing agents to work collaboratively towards a common objective. AutoDev also allows for human-in-the-loop interactions, enabling developers to provide feedback and interrupt agents as needed. The framework is designed to be integrated into IDEs, CI/CD pipelines, and PR review platforms, providing a chatbot experience for developers. AutoDev builds upon existing works in the field of AI-driven software development, including Auto-GPT, LATS, and Reflexion. It extends these works by providing a versatile tools library, empowering AI agents to autonomously perform intricate tasks, such as code editing, testing, and integration. AutoDev is also LLM-agnostic, with an infrastructure that allows a diverse set of AI models, with different parameter sizes and architectures, to collaborate on a given task. The framework has shown promising results in code and test generation, achieving high accuracy and coverage in test cases. AutoDev's design ensures a systematic and secure orchestration of AI agents to achieve complex software engineering tasks in an autonomous and user-controlled manner.AutoDev is an AI-driven software development framework that enables autonomous planning and execution of complex software engineering tasks. It allows users to define software engineering objectives, which are then assigned to autonomous AI agents. These agents can perform a variety of operations on a codebase, including file editing, retrieval, build processes, execution, testing, and git operations. AutoDev provides a secure development environment by confining all operations within Docker containers and includes guardrails to ensure user privacy and file security. The framework also allows users to define specific permitted or restricted commands and operations within AutoDev. AutoDev was evaluated on the HumanEval dataset, achieving a Pass@1 score of 91.5% for code generation and 87.8% for test generation, demonstrating its effectiveness in automating software engineering tasks while maintaining a secure and user-controlled development environment. The framework includes a Conversation Manager, a Tools library, an Agent Scheduler, and an Evaluation Environment. The Conversation Manager tracks and manages user and AI agent conversations, while the Tools Library provides a range of commands that empower agents to perform diverse operations on the repository. The Evaluation Environment allows secure execution of file editing, retrieval, build, execution, and testing commands. AutoDev's design enables autonomous AI agents to perform complex software engineering tasks without requiring developer intervention beyond setting the initial objective. The framework supports multi-agent collaboration, allowing agents to work collaboratively towards a common objective. AutoDev also allows for human-in-the-loop interactions, enabling developers to provide feedback and interrupt agents as needed. The framework is designed to be integrated into IDEs, CI/CD pipelines, and PR review platforms, providing a chatbot experience for developers. AutoDev builds upon existing works in the field of AI-driven software development, including Auto-GPT, LATS, and Reflexion. It extends these works by providing a versatile tools library, empowering AI agents to autonomously perform intricate tasks, such as code editing, testing, and integration. AutoDev is also LLM-agnostic, with an infrastructure that allows a diverse set of AI models, with different parameter sizes and architectures, to collaborate on a given task. The framework has shown promising results in code and test generation, achieving high accuracy and coverage in test cases. AutoDev's design ensures a systematic and secure orchestration of AI agents to achieve complex software engineering tasks in an autonomous and user-controlled manner.