chatgpt-export

Exporting chat conversations

repo chatgpt export

↗ GitHub

ChatGPT Conversation Exporter

Export a single ChatGPT conversation to Markdown (and optionally raw JSON) using its conversation ID.

This tool fetches the same structured JSON the ChatGPT web app uses and converts it into a clean, portable Markdown file suitable for archiving, version control, or further processing.

Features

Export a single conversation by ID
Clean Markdown output
Optional raw JSON export
Preserves roles and timestamps
No scraping — uses the same backend endpoint as the web UI

Example Output

# How to export chat

_Exported: 2026-02-25T12:34:56+00:00_

## USER (2026-02-25T12:00:00+00:00)

How do I export this chat?

## ASSISTANT (2026-02-25T12:00:02+00:00)

Here are your options...

Requirements

Python 3.10+
requests
A valid authenticated ChatGPT web session

Installation

Using UV (recommended)

uv init
uv add requests

Using pip

pip install requests

Clone the repository and place yourself in the project directory.

Usage

Basic export to Markdown:

python export_chat.py <CONVERSATION_ID> --bearer "$CHATGPT_BEARER"

Export Markdown + raw JSON:

python export_chat.py <CONVERSATION_ID> \
  --bearer "$CHATGPT_BEARER" \
  --out-md chat.md \
  --out-json chat.json

Using a session cookie instead of a Bearer token:

python export_chat.py <CONVERSATION_ID> --cookie "$CHATGPT_COOKIE"

Getting the Conversation ID

From the browser URL:

https://chatgpt.com/c/<CONVERSATION_ID>

Copy the <CONVERSATION_ID> part.

Getting Authentication

Open DevTools → Network → reload the conversation page.

Find the request:

/backend-api/conversation/<id>

From Request Headers, copy either:

Option A — Bearer token (preferred)

Authorization: Bearer <token>

Then:

export CHATGPT_BEARER="PASTE_TOKEN"

Option B — Cookie

Copy the full Cookie: header or just the session token:

export CHATGPT_COOKIE="__Secure-next-auth.session-token=..."

Then run the script.

Output Files

By default:

<conversation_id>.md

Optional:

<conversation_id>.json

Markdown includes:

Conversation title
Export timestamp
Chronologically sorted messages
Roles (user, assistant, etc.)
ISO 8601 timestamps (UTC)

How It Works

Calls:
```
GET /backend-api/conversation/<id>
```
Parses the mapping message tree
Extracts message content
Sorts by create_time
Renders Markdown

No DOM inspection. No scrolling hacks. No partial exports.

Limitations

Requires active authenticated session
API endpoint is undocumented and may change
Messages are sorted strictly by timestamp (not branch-aware)
Attachments are not downloaded (text content only)

Security Notes

Treat Bearer tokens like passwords
Never commit tokens to version control
Revoke your session if a token is exposed
JSON output may contain metadata

Suggested Archival Workflow

python export_chat.py <id>
git add <id>.md
git commit -m "Archive chat <id>"

Keeps conversations searchable and portable without using the full account export.

Possible Extensions

Branch-aware export (follow active path)
HTML output
JSONL output for indexing
Batch export multiple conversation IDs
Attachment downloading

Disclaimer

This project uses an internal backend endpoint exposed to the ChatGPT web client. It is not an officially documented public API and may change without notice. This project is based on an initial version, generated by ChatGPT.