ClaudeMap

AI Coding Assistant Data Extractor

Extract your own local chat history from ten AI coding assistants into one normalized JSONL — for fine-tuning, analytics, or backing up years of conversations. Standard library only.

SDKs & Toolingclidata-exportjsonlbackupfine-tuning
Visit projectkruzovic7/ai-data-extractor

This extractor auto-discovers and normalizes your local conversation history from ten sources — Claude Code, Codex CLI, Cursor, Windsurf, Trae, Continue, Gemini CLI, OpenCode, Cline/Roo Code and Aider — into a single JSONL format for fine-tuning, personal analytics, or plain backup before an app clears its local database. Each record keeps what the tool actually stored: user messages and assistant responses, code context (file paths, selections, snippets), diffs and suggested edits where recorded, tool calls and results, timestamps, session IDs, project paths and model names. Cursor's three known storage shapes are all implemented explicitly, undocumented Windsurf/Trae schemas are handled heuristically, and a --list flag previews what's installed without extracting anything. Python 3.9+ with zero dependencies — standard library only.

What it does

  • Ten sources enumerated and auto-discovered: Claude Code, Codex CLI, Cursor, Windsurf, Trae, Continue, Gemini CLI, OpenCode, Cline/Roo Code and Aider
  • Normalized single JSONL output keeps user messages, assistant responses, code context, diffs, tool calls and results, timestamps, session IDs and model names
  • Cursor's three known storage shapes implemented explicitly; undocumented Windsurf and Trae schemas handled heuristically
  • --list previews what's installed without extracting; --merge additionally writes one all_conversations.jsonl; zero dependencies, standard library only

When to use it

Fine-tuning datasets from your own history, personal analytics over years of agent sessions, or a one-command backup before an app update wipes its local database.

Getting started

git clone https://github.com/kruzovic7/ai-data-extractor.git
cd ai-data-extractor
python --version        # 3.9+ required, 3.10+ recommended
python extract.py       # interactive menu
python extract.py --sources cursor,claude_code,aider   # direct
python extract.py --list          # preview what's installed, extract nothing

Things to know

Local files only — it reads your own data and sends nothing anywhere, but that also means the export is only as complete as what each app actually stored. Windsurf and Trae schemas are undocumented heuristics and Cursor has changed storage shape at least three times; app updates can break extractors until the tool catches up. MIT.

Repository

kruzovic7/ai-data-extractor

Charted

Related on the map

  • Harness Bridge

    SDK / Tool

    A local launcher that points Claude Code and seven other coding harnesses at any OpenAI-, Anthropic- or Responses-compatible endpoint — without editing a single config file.

    • launcher
    • local-models
    • cli
    • menu-bar
    View

  • Claude for Siri

    SDK / Tool

    Answers macOS Spotlight and Siri prompts with your signed-in Claude Code account — an experimental macOS 27 model-delegation provider.

    • macos
    • siri
    • spotlight
    • model-delegation
    View

  • Data Maskit (数据面具)

    SDK / Tool

    A local privacy gateway for AI coding assistants — masks credentials, connection strings and PII into structured placeholders on the way out, restores them mid-stream on the way back.

    • privacy
    • gateway
    • pii
    • proxy
    View