[Feature Request] Add --include-history flag to output historical messages when resuming sessions #5135

Activity

1. github-actions

   added

   enhancementNew feature or requestNew feature or request

   area:core

   area:api

   memory

   on Aug 5, 2025
2. coygeek commented on Aug 24, 2025

   

   coygeek

   on Aug 24, 2025

   More actions

   Hey, this is a really well-thought-out feature request. I've been building some tooling around the SDK myself and ran into the exact same issue. Having to build a separate file parser just to load the history is definitely a pain point, especially since the local file format could change in the future. A unified stream from stdout would be a huge improvement for developers.

   While we wait for the team to hopefully implement something like your --include-history flag, I've been using a workaround that combines locating the transcript file with launching the process. It's not perfect, and it still has the filesystem dependency you mentioned, but it might be a bit more structured than what you're doing now.

   Based on the documentation, we know a few key things:

   • The SessionStart hook receives a transcript_path in its JSON input
   • Conversations are stored locally, and the path includes the session ID

   This means the path is predictable if you have the session ID. The integration script that's launching claude can construct the path to the transcript file itself, read it, stream the history, and then immediately start streaming the live output from the claude subprocess.

   Here’s a rough sketch of how that might look in a Python script:

   import os
   import json
   import subprocess
   import glob
   
   def find_transcript_path(session_id: str) -> str | None:
       """Finds the full path to the transcript file for a given session ID."""
       base_path = os.path.expanduser("~/.claude/projects/*")
       # Search through all project directories for the session file
       for project_dir in glob.glob(base_path):
           transcript_file = os.path.join(project_dir, f"{session_id}.jsonl")
           if os.path.exists(transcript_file):
               return transcript_file
       return None
   
   def stream_claude_session_with_history(session_id: str, prompt: str):
       """
       A workaround to stream historical messages before live messages.
       """
       print(f"--- Resuming session {session_id} with history ---")
   
       # 1. Find and stream historical messages
       transcript_path = find_transcript_path(session_id)
       if transcript_path:
           print("--- Start of Historical Messages ---")
           try:
               with open(transcript_path, 'r') as f:
                   for line in f:
                       # You can add your "isHistorical: true" flag here
                       historical_message = json.loads(line)
                       historical_message['isHistorical'] = True
                       print(json.dumps(historical_message))
           except (IOError, json.JSONDecodeError) as e:
               print(f"Error reading history: {e}")
           print("--- End of Historical Messages ---")
       else:
           print(f"Warning: Could not find history file for session {session_id}")
   
       # 2. Launch the claude process to get live messages
       command = [
           "claude",
           "--resume", session_id,
           "-p", prompt,
           "--output-format", "stream-json"
       ]
       
       print("--- Start of Live Messages ---")
       process = subprocess.Popen(command, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
   
       # 3. Stream live messages from the subprocess stdout
       if process.stdout:
           for line in iter(process.stdout.readline, ''):
               # The output is already a JSON string per line
               print(line, end='')
   
       process.wait()
       
       # Optional: print any errors
       if process.stderr:
           errors = process.stderr.read()
           if errors:
               print(f"\n--- Errors ---\n{errors}")
   
   # --- Example Usage ---
   # You would get this session ID from a previous run
   # For example, by parsing the 'session_id' from the init message.
   SESSION_TO_RESUME = "your-session-id-here" 
   CONTINUATION_PROMPT = "Okay, now based on our previous conversation, please continue."
   
   # stream_claude_session_with_history(SESSION_TO_RESUME, CONTINUATION_PROMPT)

   This still has the downsides of relying on the file format, but it centralizes the logic in one place before you kick off the live stream. Your main application logic can just consume this unified stream.

   Anyway, just wanted to share my approach in case it helps. I'm definitely giving a +1 to your feature request. It would make building robust integrations so much easier. Cheers
3. github-actions commented on Dec 7, 2025

   

   github-actionsBot

   on Dec 7, 2025 – with GitHub Actions

   More actions

   This issue has been inactive for 30 days. If the issue is still occurring, please comment to let us know. Otherwise, this issue will be automatically closed in 30 days for housekeeping purposes.
4. github-actions

   added

   autocloseIssue will be closed automaticallyIssue will be closed automatically

   on Dec 7, 2025
5. github-actions commented on Jan 7, 2026

   

   github-actionsBot

   on Jan 7, 2026 – with GitHub Actions

   More actions

   This issue has been automatically closed due to 60 days of inactivity. If you're still experiencing this issue, please open a new issue with updated information.
6. github-actions

   closed this as not plannedon Jan 7, 2026
7. github-actions commented on Jan 14, 2026

   

   github-actionsBot

   on Jan 14, 2026 – with GitHub Actions

   More actions

   This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.
8. github-actions

   locked as resolved and limited conversation to collaborators on Jan 14, 2026