If you’ve been using Claude Code to build something real, I have bad news. Every conversation you’ve had with it—every prompt, every bug fix, every architectural decision—has an expiration date you probably didn’t know about.
Default settings are design decisions. And this one says your work is disposable.
Claude Code’s out-of-the-box configuration automatically deletes your conversation history after 30 days. You can change it by tweaking `cleanupPeriodDays` in your settings.json—but only if you know it exists. Most people don’t. They discover the loss only when they go back to revisit something they discussed weeks ago, only to find an empty slate.
This isn’t a bug. It’s a philosophy.
The tool is marketed as an AI coding assistant that builds long-term understanding. It learns from your patterns, remembers your preferences, and accumulates context. Yet its default behavior treats that context as temporary—like a sticky note you’re expected to throw away after a month.
The real problem isn’t the 30-day limit. It’s that the tool treats you like a tenant, not an owner.
One user on Hacker News put it bluntly: “Their interface aggressively resists the idea of saving the answer for future consumption. No, I do not need that incredible slow PDF builder, just give me to download HTML.” Another noted that even when transcripts survive, after days of compact compression, the system “fails to find conversation or important things discussed.”
So you’re not just losing history after 30 days. You’re losing the ability to reliably search through the history you do have.
This is the silent assumption that you don’t own your transcripts—the platform does.
Anthropic built a powerful tool. But with great power comes great responsibility to let users keep what they’ve created. Every other major AI tool (ChatGPT, Gemini, Copilot) gives you control over export and retention by default. Claude Code’s choice to hide this behind a settings.json flag is a deliberate product decision that prioritizes system efficiency over user ownership.
I’ve seen firsthand what happens when this catches you off guard. Hours of troubleshooting, design rationale, and incremental progress vanish. The conversations weren’t just logs—they were the bridge between your last session and your next insight. Losing them means re‑asking questions, re‑explaining context, and re‑building trust.
If you rely on Claude Code for coding or research, your entire conversation history is at risk of automatic deletion unless you change that setting right now.
Open your settings.json. Find `cleanupPeriodDays`. Set it to a value that matches your real workflow—365, 999, or remove the limit entirely. Then ask yourself: Why was this option buried in the first place?
The 30-day default isn’t a bug. It’s a statement. And it’s time we demanded better.
FAQ
Q: How do I prevent Claude Code from deleting my conversations?
A: Open your settings.json file (usually in the Claude Code config directory) and set `cleanupPeriodDays` to a value like 365 or `null` to disable automatic deletion. Then save and restart Claude Code.
Q: Why doesn't Claude Code offer a simple export option?
A: The interface intentionally resists saving transcripts—it offers a slow PDF builder instead of a plain HTML download. This is a design choice that favors platform control over user ownership. You can manually copy conversations or use third-party scripts, but Anthropic hasn't prioritized export.
Q: Is this a privacy concern or just a UX flaw?
A: Both. The 30-day default assumes you don't need long-term access, which contradicts the promise of a tool that 'remembers' context. It also means Anthropic can claim they don't store old data—but at the cost of your ability to revisit your own work. The lack of transparency is the real issue.