Configuration file for Google Jules AI coding agent, providing project-specific instructions, architecture context, and coding standards.
JULES.md
JULES.md is a project-level configuration and instructions file read by Google Jules, an autonomous AI coding agent, to understand a repository’s architecture, conventions, and coding standards before it makes changes. Placed at the root of a repository, it plays the same role for Jules that AGENTS.md or CLAUDE.md play for other coding agents — a durable, human-authored brief the agent loads as context.
- Project context - Describes the architecture, key directories, and how the codebase fits together.
- Coding standards - Encodes the style, testing, and contribution conventions the agent should follow.
- Task guidance - Tells the agent how to build, test, and verify its work in the repository.
- Version-controlled - Lives in the repo alongside the code, so the guidance evolves with the project.
In practice JULES.md is committed to a repository so every Jules run starts from the same shared understanding, reducing the back-and-forth of re-explaining a project to the agent on each task. It is part of a broader shift toward agent-readable instruction files that make codebases — and, increasingly, API operations — legible to autonomous agents, the same instinct behind treating OpenAPI, AGENTS.md, and llms.txt as first-class context for machines.