WhyChose

Topic: how to export claude conversations

How to Export Your Claude Conversations

Anthropic's export is simpler than OpenAI's — flatter JSON, faster turnaround, no branching DAG — but it's buried in a different menu. Here's the exact path and what the archive looks like.

TL;DR

Go to claude.ai → click your initials (top-right) → Settings → Account → scroll to Export data → Request. Anthropic emails a ZIP link within 15–30 minutes. The archive contains conversations.json (flat array, one object per conversation, messages as a simple list) and users.json. No HTML viewer, no image blobs — just JSON.

Why this matters

If you work across both Claude and ChatGPT (as most senior engineers do in 2026 — ChatGPT for quick lookups, Claude for long-form reasoning and code), you'll want both archives when you audit your thinking. Claude export is the simpler of the two to parse, which makes it a good starting point for anyone building a decision-extraction pipeline or personal chat search. The JSON is stable — Anthropic hasn't changed the shape since late 2024.

How to approach it

  1. Open claude.ai in a browser. The mobile app doesn't expose this setting as of April 2026.
  2. Click your initials in the top-right corner → Settings.
  3. Go to the Account tab in Settings.
  4. Scroll to "Export data" near the bottom → click Request. A confirmation modal appears; confirm.
  5. Watch your inbox. The email comes from no-reply@anthropic.com with subject "Your Claude data export is ready." Download within 7 days — Anthropic's link TTL is longer than OpenAI's.

What's inside the ZIP

Two files:

conversations.json    ← all your chats, flat array
users.json            ← account metadata

Each conversation object is refreshingly flat compared to ChatGPT:

{
  "uuid": "abc-123",
  "name": "Postgres vs MongoDB for metrics",
  "created_at": "2026-02-14T10:04:22Z",
  "updated_at": "2026-02-14T10:22:41Z",
  "chat_messages": [
    { "uuid": "...", "sender": "human",     "text": "..." },
    { "uuid": "...", "sender": "assistant", "text": "..." }
  ]
}

Key difference vs ChatGPT: Claude gives you chat_messages[] as a linear array. ChatGPT gives you a mapping DAG that branches on message edits. For Claude you just iterate; for ChatGPT you have to flatten the tree. If you're writing code that consumes both, handle the two shapes with different loaders and emit a common internal format.

How WhyChose helps

WhyChose detects the export format automatically. Drop a Claude conversations.json or a ChatGPT one into the uploader and it routes to the right loader. The open-source CLI supports both — run node bin/extractor.js path/to/conversations.json --format md and you get a decision log back in either case. Decision records are emitted in the same shape regardless of source, so your downstream audit trail is platform-agnostic.

Get early access

Related questions

Does the export include Projects and Artifacts?

Yes. Projects appear as conversations with an additional project_uuid field. Artifacts (Claude's structured code/document outputs) are preserved inline in the assistant message as part of the message text, not as separate blobs.

Is there an admin export for Claude Team?

Yes — Team and Enterprise admins can export all workspace conversations via the admin dashboard's Data Management section. Individual-user exports only return that user's own chats.

Can I export from the API instead of the UI?

Not for historical conversations. The Anthropic Messages API is stateless — it doesn't retain history, so there's nothing to export from it. The ZIP flow above is the only way to get your existing chat history out.

How often can I re-request an export?

Anthropic doesn't publish a hard rate limit, but in practice you can request every few hours. For a quarterly decision-audit workflow, requesting once every 90 days is more than enough.

Further reading