Codex Cannot Access Its Local Database — How to Fix It

Just updated Codex Desktop and now it won't open? You're seeing this scary message:

Codex cannot access its local database. The app cannot finish launching until its SQLite database is accessible. Most recent error: migration 1 was previously applied but has been modified

Don't worry. Your work is safe. This is a known bug, and the fix takes about two minutes.

Here's everything you need to know — the quick fix first, then the why.

Quick Fix (2 Minutes)

This is the fastest fix and it does not delete your chats or login.

1. Fully close Codex. Quit the app completely. Check your system tray (bottom-right on Windows) and close it from there too.
2. Open your .codex folder. Go to this path:
   • Windows: C:\Users\<your-username>\.codex\
   • Mac / Linux: ~/.codex/
3. Find these two files and rename them (just add .old to the end so you have a backup):
   • state_5.sqlite → state_5.sqlite.old
   • logs_2.sqlite → logs_2.sqlite.old
4. Restart Codex.

That's it. Codex will rebuild those two files automatically and open normally.

Important: Only touch those two .sqlite files. Do NOT delete the sessions folder or the auth files inside .codex — that's where your chat history and login are stored.

Why This Error Happens

This error started showing up after the Codex Desktop 0.130 → 0.131 update (around May 2026). The auto-update changed something small inside the app's database files, and it no longer matched what was already saved on your computer.

In plain English: Codex keeps a small local database to track your settings and history. The update slightly changed the "blueprint" of that database, but your old file still followed the old blueprint. When the two don't match, Codex refuses to open — just to be safe.

The technical name for this is a migration mismatch (a "CRLF/LF checksum drift," if you want the nerdy version). It mostly hits Windows and WSL users, but Mac and Linux users have seen it too.

The good news: the two files causing the problem (state_5.sqlite and logs_2.sqlite) only hold metadata and indexing info. Your real conversations live in a separate sessions folder that this bug never touches. That's why deleting or renaming those two files is totally safe.

Read Now: OpenAI Codex Just Replaced My Junior Developer (Here's How)

What "Migration 1 Was Previously Applied But Has Been Modified" Actually Means

This is the part of the error that scares people the most, so let's break it down in plain English.

Codex keeps a small database on your computer to remember your settings, recent activity, and how things are organised. Every version of Codex expects that database to follow a certain "blueprint."

When you updated the app, the new version came with a slightly changed blueprint. But your old database file on disk still followed the old blueprint. So when Codex started up, it compared the two, saw they didn't match, and basically said: "This doesn't look right — I'm not opening until someone fixes it."

That mismatch is exactly what the message "migration 1 was previously applied but has been modified" is describing. A "migration" is just the technical word for updating a database's structure. Codex is telling you that a migration it already ran in the past no longer matches what it expects now.

The deeper cause is a CRLF/LF checksum drift — a tiny difference in how line endings got saved in the state_5.sqlite and logs_2.sqlite files. You don't need to understand that part. You just need to know it's harmless and easy to clear with the fix above.

If the Quick Fix Doesn't Work

Sometimes you'll see a slightly different version of the error, like:

• attempt to write a readonly database
• database disk image is malformed
• timed out waiting for state db backfill after 30s

Try these steps in order:

Step 1: Check if the file is corrupted

Open a terminal and run:

bash

sqlite3 ~/.codex/state_5.sqlite "pragma integrity_check;"

If it says anything other than ok, the file is damaged — renaming it (Quick Fix above) will solve it.

Step 2: Use the Repair / Reset option

The Codex app has a built-in Repair and Reset option. Try Repair first. If that fails, use Reset (this clears local data but keeps your account).

Step 3: Reinstall the app

If nothing works:

1. Uninstall Codex
2. Delete the .codex folder (back it up first if you want to keep old chats)
3. Reinstall Codex from the official source
4. Sign in again

Step 4: Use a repair toolkit

A developer has published a free open-source tool called codex-repair on GitHub that fixes this exact crash automatically — without losing any conversations. It only touches the two problem files and leaves your chat history read-only the entire time. Search "xdifu codex-repair" on GitHub if you want the automated route.

Will I Lose My Chats or Login?

No. This is the most common worry, so let's be clear:

• Your chat history is stored in ~/.codex/sessions/ — the fix never touches it.
• Your login / auth is stored separately — you stay signed in.
• The only thing that gets rebuilt is the small metadata database, which Codex recreates fresh on the next launch.

The worst case is that your recent thread list takes a moment to re-index. Everything else stays exactly where it was.

How to Avoid This in the Future

This bug is tied to a specific bad update, and OpenAI is rolling out fixes. To stay safe:

1. Back up your .codex folder before any major Codex update.
2. Don't run the Windows app and WSL CLI on the same database at the same time — that's a common trigger for this error.
3. Keep Codex updated, since the team is patching these database issues quickly.

Read Now: GPT-5.1: Everything You Need to Know

Final Word

The "Codex cannot access its local database" error looks scary, but it's almost always harmless. Nine times out of ten, renaming state_5.sqlite and logs_2.sqlite and restarting the app fixes it in under two minutes — with zero data loss.

If you hit a more stubborn version of the error, the Repair tool or a clean reinstall will get you back up and running.

Now get back to coding.

FAQs