How to search your Claude chat history

Claude now has several retrieval paths, but availability and project boundaries matter. Here is the fastest option for each situation.

By Niko Nishimine · Updated July 25, 2026

Option 1: Use Claude’s Chats history

Open Chats from the left sidebar to browse past conversations. Renaming important chats as you work makes this much more useful. If you already know the conversation, open it and use Cmd-F on Mac to search the text currently loaded on the page.

Option 2: Ask Claude to search past chats

Anthropic is rolling out a feature that lets Claude search and reference previous conversations when you ask naturally—for example, “What did we decide about the onboarding flow?” Anthropic currently documents this for Max, Team, and Enterprise users, with expansion planned.

There are important boundaries:

• Chats outside projects can be searched together.
• Project conversations are searched within that individual project.
• The feature can be turned off in profile preferences.
• Availability may differ by plan and rollout status.

Option 3: Use Projects deliberately

Claude Projects group chat histories and a knowledge base. They are excellent when you already know which project contains the answer. Anthropic also notes that context is not automatically shared across separate project chats unless the relevant information is added to project knowledge, so organization and active context are not quite the same thing.

Option 4: Export your Claude data

For a complete backup, go to Settings → Privacy → Export data on the web or Claude Desktop. Anthropic’s data-export guide says the export includes conversation and account data and arrives through an emailed download link. This is useful for archival research, but it is a snapshot rather than a continuously updated search tool.

When local full-text search is useful

A local index is useful if you want predictable keyword search, highlighted matches, and immediate access regardless of which provider feature is available to your plan. It also keeps retrieval separate from model personalization: searching your archive does not require asking the model to infer which old discussion you meant.