Index your Claude export in ModelBrain

• Request your export from Claude in its privacy settings and download it. The download is a set of zips with a manifest. The one ModelBrain reads is conversations-000.zip, which holds conversations.json.
• Use ModelBrain 0.6.6 or later. You can use the zip as it is, a folder that holds the zips, or an unzipped conversations.json.
• Nothing is added until you add it. ModelBrain never guesses a source kind from a file extension.

1. Open the ModelBrain menu, choose Vaults, then Add Source, then Claude export….
2. A prompt asks whether your export is a zip file. Choose Yes to pick the zip. Choose No to pick the folder.
3. ModelBrain indexes it in the background. A large export takes a while.

From a terminal, connector_sync register <id> <path> --type claude-export does the same. For example: connector_sync register claude ~/Downloads/conversations-000.zip --type claude-export.

• The conversations.json file, streamed. It is never held whole in memory.
• One memory per conversation, keyed by its uuid. The title is the conversation name and the date is when it was created.
• The text of your messages and Claude's replies. A message that has files and no text becomes [file: name], or [attachment]. No attachment contents are read.
• Each turn sits under a User or Assistant heading, so search can tell who said what. A passage you wrote ranks above the same words in a reply.
• If two exports overlap, the copy with the later update time wins.

• Thinking, tool use and tool results, and prompts injected by apps.
• Attachment bodies.
• Conversations with no dialogue text. They are skipped and counted.
• Side branches. The export marks no branch, so ModelBrain takes the chain from the newest message back through the parent links. A conversation with no usable parent links keeps every message in time order.
• The other zips in the export (memories, projects, design chats and the rest). This source never indexes them.

A container file over 4 GiB is refused and recorded as a failure, not dropped. Set MODELBRAIN_CONTAINER_MAX_BYTES to change that ceiling. A single conversation over 32 MiB is refused the same way, and the rest of the export carries on.

While it indexes, ModelBrain writes the conversation text to a private temporary file, then deletes it when the work is done. Allow a few hundred megabytes of free disk space for a large export. If the export cannot be read to the end, every conversation already indexed stays put until the next good pass.

1. Ask your assistant to run ModelBrain diagnostics. The source reports its kind, members, members_failed and last_container_bytes.
2. Ask a local assistant about something you once discussed with Claude.