copilotexport

Export all of your VS Code GitHub Copilot Chat conversations — including full tool-call history — to JSON and Markdown. Mirrors the layout of ChatGPT and Claude exports.

Installation

pip install copilotexport

Usage

CLI

# Default: bundle every Copilot session into ./CopilotForAGiXT.zip,
# ready to upload to AGiXT (Conversations sidebar -> Import).
copilotexport

# Write the zip somewhere else
copilotexport --out ~/CopilotForAGiXT.zip

# Custom source workspaceStorage directory
copilotexport --src /path/to/Code/User/workspaceStorage

# Full human-browseable export tree (raw JSON + Markdown + index +
# CopilotExport.zip) — slower, useful for grepping / archiving.
copilotexport --full
copilotexport --full --out ~/my-export
copilotexport --full --no-markdown   # skip Markdown rendering (much faster)
copilotexport --full --no-zip        # skip CopilotExport.zip
copilotexport --full --no-raw        # skip raw VS Code JSON copies

Importing into AGiXT

The default mode writes a zip containing a single conversations.json — a JSON array of raw VS Code session dicts. Upload it through the Import Conversations control in the AGiXT web UI, or POST it directly:

curl -F file=@CopilotForAGiXT.zip -F agent_name=XT \
     -H "Authorization: $JWT" \
     http://localhost:7437/v1/conversation/import

AGiXT detects the format as copilot, prefixes imported conversations with [Copilot], preserves each conversation's original timestamps, and renders Copilot tool invocations / thinking blocks as [SUBACTIVITY] entries in the chat UI.

Python API

from copilotexport import export, default_workspace_storage
from pathlib import Path

summary = export(
    src=default_workspace_storage(),
    out=Path("_copilot"),
    write_markdown=True,
    make_zip=True,
    copy_raw=True,
)
print(summary)

Output layout

_copilot/
├── export_info.json                         # run summary
├── index.json                               # flat metadata for all sessions
├── workspaces.json                          # wsHash → folder URI map
├── sessions/
│   └── <workspace_label>/
│       └── <sessionId>.json                 # verbatim VS Code session file
└── markdown/
    └── <workspace_label>/
        └── <YYYY-MM-DD>_<title>_<id8>.md   # human-readable transcript

Each Markdown transcript includes:

• Session metadata (title, session ID, timestamps, model, agent)
• User messages
• Assistant responses with prose rendered inline
• Tool calls rendered as structured blocks showing the tool name, request summary, result summary, and any file paths touched

Platform support

VS Code's workspaceStorage is auto-detected on Linux, macOS, and Windows.