Claude Code /heapdump: Diagnose Memory When AI Keeps Slowing Down

After working for a long time, there comes a moment when Claude Code feels sluggish compared to the start. Input responses slow down, the fan spins up, and something feels stuck. At times like this, instead of guessing "why did it slow down?", /heapdump lets you capture and inspect the actual memory state.

Definition (What It Is)

/heapdump is a command that saves to a file the snapshot of how Claude Code is using memory at this moment.

Here, the heap is the memory space where a program stacks up data while running. A dump means exporting the contents of that space wholesale to a file. So a heapdump can be thought of as "a snapshot of what and how much is piled up in memory right now."

Just as taking a photo lets you examine it slowly later, capturing a memory snapshot lets you analyze what is eating up your memory.

How to Use It (By Difficulty)

Basic — Taking One Snapshot

When Claude Code feels heavy, just type it.

/heapdump

This saves the current memory state as a snapshot file. No options needed — one line does it.

Intermediate — Analyzing the Captured File

Saving a snapshot is easy, but reading the file requires a separate tool. Heap snapshots are usually saved in a format you can open in the Memory tab of Chrome DevTools.

The analysis flow goes like this.

1. Run /heapdump at the moment it slows down to save a snapshot.
2. Check the saved file path (it is shown on screen after running).
3. Open the Chrome browser, go to DevTools (F12) → Memory tab, and load that file.
4. Check the list of which data takes up the most memory.

Advanced — Comparing Across Points in Time

If you suspect a memory leak (a phenomenon where unused data keeps piling up and memory "leaks"), it is effective to take two snapshots and compare them.

/heapdump

Take one right after starting work, and another after working for a while until it slows down. Comparing the two snapshots reveals which data has grown abnormally over time. Normally things should stay at a steady level, so if there's an item that keeps growing, it is likely the culprit.

Common Pitfalls

• Snapshot files are larger than you'd expect. They can range from tens to hundreds of MB, so it's good to delete them once analysis is done.
• /heapdump is only a diagnostic tool; it does not reduce memory or fix the problem. Think of it as the first step in finding the cause.
• The moment you take a snapshot, it may briefly hitch. That's because it reads all of memory and writes it to a file. It's normal, so don't be alarmed.

Real-World Example

A beginner developer ran Claude Code for several hours on a large project and noticed input becoming noticeably slow. Restarting helped briefly, but the slowdown kept recurring. They took /heapdump twice — early in the work and after it slowed down — and comparing them revealed that certain data kept accumulating. Switching to the habit of breaking heavy tasks into smaller pieces and starting fresh sessions made the problem disappear.

Taking It Further

If you need memory diagnostics, there are commands worth using around it. To check whether the environment itself has issues, use /doctor to verify the Claude Code installation, and if you want to see more debugging information, look into the /debug family of commands. Once /heapdump confirms "memory is the problem," the next step is to figure out how to make your workflow lighter.

Summary

/heapdump is a diagnostic command that captures the memory state like a photo when Claude Code gets heavy. Taking it once shows the current state; taking it twice and comparing tells you whether memory is leaking. Reach for it when you want to judge from real data rather than guesswork.

Based on: Claude Code v2.1.154 (2026.05)