Deep research has emerged as a leading application in the AI agent space, and the Open Deep Research project is a fully open-source agent that integrates with various model providers and search tools. Its performance ranks high among existing deep research agents, evidenced by its #6 position on the Deep Research Bench leaderboard. The project has seen regular updates, including the integration of GPT-5 and a free course on building open deep research agents. To get started with Open Deep Research, users can clone the repository and set up a virtual environment. The installation process involves syncing dependencies and configuring environment variables for model selection and search tools. The agent can then be launched with the LangGraph server, allowing users to interact through a web interface. The agent employs several large language models (LLMs) for tasks like summarization, research, and report writing, with default configurations using OpenAI's models. The Deep Research Bench evaluation framework consists of 100 PhD-level research tasks across multiple fields, designed to reflect real-world research needs. Running evaluations can be costly, ranging from $20 to $100, depending on the chosen model. The results can be exported and submitted to the Deep Research Bench for performance assessment. Open Deep Research also supports various MCP tools and search APIs, allowing users to customize their agent setup based on specific research challenges. For non-technical users, the Open Agent Platform (OAP) provides an accessible interface to build and configure agents like Deep Researcher without coding. Users can test the agent on a public demo instance by simply adding their API keys. The project also includes earlier implementations that explore different approaches to automated research, such as structured workflows and parallel processing, though these are less efficient than the current version.