Chatlog

<project-slug> = working directory absolute path with / replaced by - (e.g., /home/ubuntu/wb → -home-ubuntu-wb).

Find current session: ls -t ~/.claude/projects/<project-slug>/*.jsonl | head -1

JSONL format:

• Each line is a JSON object with "type" field being "user", "assistant", or "file-history-snapshot"
• User messages: record.message.content is a plain string
• Assistant messages: record.message.content is a list of items, extract only items where item.type == "text" and take item.text
• Consecutive same-role messages should be merged (streaming chunks)
• Skip records where type is not "user" or "assistant"

Inline extraction script — you can run it directly:

import json, sys
from pathlib import Path

def extract_claude(input_path, output_path):
    messages = []
    with open(input_path, "r", encoding="utf-8") as f:
        for raw in f:
            line = raw.strip()
            if not line:
                continue
            rec = json.loads(line)
            if rec.get("type") not in ("user", "assistant"):
                continue
            msg = rec.get("message", {})
            role = msg.get("role")
            if role not in ("user", "assistant"):
                continue
            content = msg.get("content")
            if isinstance(content, str):
                text = content.strip()
            elif isinstance(content, list):
                text = "\n".join(
                    item["text"] for item in content
                    if item.get("type") == "text" and isinstance(item.get("text"), str)
                ).strip()
            else:
                continue
            if not text:
                continue
            if messages and messages[-1]["role"] == role:
                messages[-1]["content"] += "\n" + text
            else:
                messages.append({"role": role, "content": text})
    Path(output_path).write_text(
        json.dumps({"messages": messages}, ensure_ascii=False, indent=2) + "\n", encoding="utf-8"
    )
    print(f"Extracted {len(messages)} messages -> {output_path}")

if __name__ == "__main__":
    extract_claude(sys.argv[1], sys.argv[2])

Codex

Session files location:

~/.codex/sessions/YYYY/MM/DD/rollout-<timestamp>-<uuid>.jsonl

Find current session: find ~/.codex/sessions -name '*.jsonl' -printf '%T@ %p\n' | sort -rn | head -1 | cut -d' ' -f2

JSONL format:

• Filter lines where record.type == "response_item" and record.payload.type == "message"
• Role is at record.payload.role, keep only "user" and "assistant"
• Content is at record.payload.content (a list), extract items where item.type is one of "input_text", "output_text", "text" and take item.text

Inline extraction script — you can run it directly:

import json, sys
from pathlib import Path

TEXT_TYPES = {"input_text", "output_text", "text"}

def extract_codex(input_path, output_path):
    messages = []
    with open(input_path, "r", encoding="utf-8") as f:
        for raw in f:
            line = raw.strip()
            if not line:
                continue
            rec = json.loads(line)
            if rec.get("type") != "response_item":
                continue
            payload = rec.get("payload", {})
            if payload.get("type") != "message":
                continue
            role = payload.get("role")
            if role not in ("user", "assistant"):
                continue
            content = payload.get("content", [])
            if not isinstance(content, list):
                continue
            text = "\n".join(
                item["text"] for item in content
                if item.get("type") in TEXT_TYPES and isinstance(item.get("text"), str)
            )
            if text:
                messages.append({"role": role, "content": text})
    Path(output_path).write_text(
        json.dumps({"messages": messages}, ensure_ascii=False, indent=2) + "\n", encoding="utf-8"
    )
    print(f"Extracted {len(messages)} messages -> {output_path}")

if __name__ == "__main__":
    extract_codex(sys.argv[1], sys.argv[2])

Step 3: Execute

1. Find the current session jsonl file using the method above
2. You can run the extraction code directly.If it doesn't work, you should write the appropriate extraction script to a temp file (e.g., /tmp/_chatlog_extract.py) and then run it.
3. If you want to verify the output file, do it after the extraction command finishes. Do not read chatlog.json in parallel with the write, or you may hit a race and see stale/missing output.
4. Report the number of extracted messages and the output path to the user