Does Claude remember previous conversations or does it start fresh every time?

By default Claude does not have persistent memory between separate conversations. Each new conversation starts completely fresh with no knowledge of what you discussed before. This is a fundamental aspect of how large language models work — the model only knows what is inside the current context window of the active conversation.

Claude does have a memory feature in claude.ai that can be enabled. When turned on, Claude will save certain facts you share about yourself across conversations, like your job, preferences, or recurring context. You can view and manage these saved memories in your settings. This feature is optional and you stay in control of what gets stored.

For developers building applications with the Claude API, memory is something you implement yourself. The common patterns are: storing conversation history in a database and passing the relevant portions with each new API call, using a vector database to retrieve semantically relevant past exchanges, or summarizing past context and injecting the summary into the system prompt. Libraries like LangChain and LlamaIndex have built-in utilities to handle this for Claude.