The .claude Folder: How Claude Stores Local State in Your Project

📁 What Is the `.claude` Folder and What Does It Do?
#

If you’ve ever opened a project and found a .claude folder you didn’t create, you’re not alone. This folder is automatically created by tools that integrate Anthropic’s Claude, and it plays a central role in how the model operates within your project.

🗂️ What’s Inside?
#

config.json – Model behavior configuration: which model to use, constraints, and settings.
memory/ or context/ – Persistent context between sessions. This is how Claude “remembers” previous interactions.
agents/ or tasks/ – Workflow definitions and multi-step agent logic.
logs/ – Execution history, errors, and results.
cache/ – Temporary data to speed up future runs.

⚙️ How Does It Work?
#

Every time you run a task, the system:

Reads config.json to understand how to behave.
Loads stored context from memory/.
Executes the task (simple or multi-step).
Writes logs, updates context and cache.

Each run builds on the previous one — that’s what gives continuity to AI-powered workflows.

⚠️ Common Mistakes
#

🚫 Deleting it without understanding what you lose – Resets all memory and project configuration. 🔓 Committing it to git – Can expose sensitive data like logs with user inputs. 📋 Ignoring the logs – They’re the first diagnostic source when something goes wrong.

💡 In Simple Terms
#

The .claude folder is like the “working memory” of Claude in your project. Without it, every interaction starts from scratch. With it, the system maintains continuity, knows how to behave, and can run complex workflows consistently.