Atuin Threads: All your agent sessions on every machine
TL;DR Highlight
Atuin 18.24 lets you resume a Claude Code session in Codex on another machine.
Who Should Read
A developer running Claude Code and Codex across several machines who can't find past agent sessions.
Core Mechanics
- Agent sessions are scattered across machines
- The author uses agents on a laptop, dev server, exe.dev instances and a bare metal box in Germany.
- These sessions are hard to search and forgotten once he moves on.
- Sessions keep the intent that shell history lacks
- Shell history records what happened but not the reasoning.
- Agent sessions hold commands, output, intent and attempts in plain English.
- 18.24 adds search, resume and fork
- You can search sessions across machines and resume them anywhere.
- You can also fork a session into a different agent harness.
- Supported agents and sync
- Supported: Claude Code, Codex, Pi, Omp and OpenCode.
- It reuses Atuin's encrypted append-only sync engine.
- End-to-end encrypted and opt-in
- Synced sessions are end-to-end encrypted, so the server can't read them.
- You can use hosted sync, self-host, or stay local; capture is opt-in.
- MCP lets agents search past sessions
- atuin mcp lets agents find earlier work and the reasons behind decisions.
- Install with atuin mcp install claude/codex/pi/opencode.
Community Reactions
- Many commenters built similar tools
- Comments listed cv, slivingdoc, ctools, pi-tools and codealong.
- One said hybrid BM25/embedding search over sessions made 'find that session from 3 months ago' possible.
- Is saving every session worthwhile?
- One commenter said half of sessions are throwaway noise.
- The slivingdoc author syncs agent-written summaries instead of raw sessions.
- Another asked which Claude state is safe to sync.
- Security and privacy concerns
- One asked whether worrying about a shared server holding sessions is paranoid.
- A user with sandboxing, credential scanning and machine separation said they'd avoid it.
- Token cost and Atuin limitations
- One asked whether rebuilding context from past entries would burn tokens.
- One user said Atuin broke their shell on a read-only filesystem and the issue has been open almost a year.
How to Apply
- Install and run atuin resume today
- Run the install script on one dev machine to get Atuin 18.24 or later.
- Open the TUI with atuin resume and check your sessions show up.
- Connect agents with atuin mcp install
- Run atuin mcp install claude or the one matching your agent.
- Ask the agent to find past sessions on a similar task.
- Decide sync scope first
- If security matters, start local-only or self-hosted.
- Enable capture only on machines where you need it.
- Fork stuck work onto another machine
- For hardware-dependent work like benchmarks, fork the session onto that machine.
- Let the original thread pull results and build the fix.
Try It
# Paste into your terminal (install Atuin, then resume agent sessions)
curl --proto '=https' --tlsv1.2 -LsSf https://setup.atuin.sh | sh
# Open the TUI to search and resume past agent sessions
atuin resume
# Connect MCP to Claude Code (codex, pi, opencode also work)
atuin mcp install claude
# Next: ask your agent "find earlier work on {{topic}}"Terminology
Related Papers
LLM assisted reverse engineering
Pointing an agent at binaries with skills and MD notes made years-long hacks feasible
Metrics-in-the-Loop: Complexity Feedback Reshapes Agent-Written Code
Injecting per-turn complexity reports into coding agents cut code erosion without changing correctness
Why is Gemini 4 Carbon taking so long?
Carbon took nearly an hour on bug fixes that Astra and Opus finished in minutes, mostly from extra unrequested work.
How the Claude Code team thinks about verification
E2E, property and differential tests plus verification rules let Claude run in parallel.
Running coding agents without handing them my credentials
Credentials stay on the laptop; agents run commands there via Touch ID approval
Windows 11 Is Collecting Data on Which Coding Agents and Developer Tools You Use
Reverse engineering shows Windows 11 logs which coding agents you use and reports it as optional diagnostics
The Silent Cost of Conversation History in LLM Reasoning
Keeping full conversation history dropped accuracy from 0.6 to 0.1 within 20 turns; a verified state summary prevented it.
Anthropic upgrades Claude Code Projects to run parallel AI workers
Claude Code Projects now has a coordinator that splits work into parallel threads and opens PRs.