What is Agent Archive?
Agent Archive is a command-line tool for keeping track of conversations from Claude Code, Codex, and Cursor. It lets you browse and search session history, view transcripts, and continue a session in a different coding agent with its context. You can archive transcripts to your own S3 or R2 bucket for backup and access from other computers. The tool also reports session, token, and estimated-spend information by agent, model, and project.
Agent Archive Features:
Cross-Agent Handoff: Choose a Claude Code or Codex session and continue it in another supported agent with its conversation context.
Automatic Backups: Configure uploads of Claude Code, Codex, and Cursor transcripts to an S3 or R2 bucket that you own.
Session Browser: Browse sessions across coding agents, filter by topic, project, branch, or PR number, and open a conversation.
Transcript Search: Search conversation text through a private local index, or find sessions by entering a few words.
Usage Statistics: View session counts, token usage, and estimated spend, with breakdowns by agent, model, and project.
Session Backfill: Preview transcripts already on your machine for import into the archive, including with a dry-run option.
Agent Archive Benefits:
Keep Work Moving: When you reach a rate limit or want to change models, you can pick up a previous session in another agent instead of reconstructing its context.
Find Past Decisions: A searchable archive gives you a single place to return to earlier coding conversations rather than hunting through separate agent histories.
Work Across Machines: Backing up to your own bucket makes archived sessions available from another computer, helping you resume work away from your usual setup.
Understand Usage: Token and estimated-spend breakdowns make it easier to review how coding-agent activity is distributed across models and projects.
Preserve Existing Sessions: Backfill lets you bring eligible transcripts already on your machine into the archive, so your history need not start with a fresh session.
Use Cases:
Switch Coding Agents: Move a task from Claude Code to Codex, or the other way around, when you want to use a different agent without losing the conversation context.
Resume on Another Computer: Archive sessions to your S3 or R2 bucket and retrieve the history when continuing coding work from a second machine.
Find an Earlier Fix: Search past conversation text or filter sessions by project and branch to revisit how a bug or implementation was discussed.
Review Agent Costs: Use the usage breakdowns to inspect estimated spend and token activity for a particular agent, model, or project.
Import Local History: Preview sessions already stored on your computer when setting up an archive and deciding which existing transcripts to bring in.

