Sources

Claude

Ingest your Claude (Anthropic) export ZIP — every conversation and message.

The claude source ingests the data export ZIP from claude.ai. Each conversation becomes a parent item, each message a child — the same shape as the chatgpt source.

Get the export

Sign in at claude.ai.
Click your profile picture → Settings → Privacy → Export data.
Anthropic emails a download link when the archive is ready.
Download the ZIP locally.

Register the export

bash

vault add-export \
  --source claude \
  --path ~/Downloads/claude-export.zip \
  --account myname@example.com

Use the email you sign into Claude with — it lets you ingest multiple Claude accounts (work, personal) into one vault without collision.

What gets ingested

Item	Notes
One conversation item per chat	Title, created/updated timestamps, model used.
One message item per turn	`kind=message`, role (`user` / `assistant`), text body, ts.
Project metadata	When a conversation was inside a Claude Project, the project name lands in metadata + tags.
Tool calls	Preserved as messages with the tool's role.

Ingest it

bash

vault ingest --source claude

vault

$ vault ingest --source claude
Extracting claude-export.zip ...
Parsing conversations (812 threads) ...
  conversations: 812
  messages:     11,401
Run 12 completed in 4.9s

Caveats

No file attachments. Files you uploaded into Claude conversations aren't in the export. Only the text Claude saw is preserved.
No artifacts. Generated artifact contents are captured as text in the message, but rendered output / interactive artifacts aren't separately preserved.
Project knowledge. Project-level knowledge documents may or may not be in the export depending on Anthropic's current export shape — when present they're ingested as conversation context.