Claude Code Suddenly Slow and Dumb? Check /context First

Has Claude Code, which was working fine just a moment ago, suddenly gotten slow, forgotten what you just said, and started giving vague answers? Nine times out of ten, its memory is full. People can't focus when their head is cluttered either. /context is a command that shows that head as a picture. When it feels slow, turn this on first.

Definition

/context is a command that shows, as a colored grid picture, how much memory (context) your conversation is currently using.

Here, "context" is the amount of information AI can hold in mind at once in a single conversation. Think of it as a bowl of fixed size. The more files you have it read and the longer the conversation, the fuller the bowl, and when it's full, older content gets pushed out, making answers fuzzy and slow. /context tells you how full this bowl is now, and even what's taking up a lot of space.

How to use it (by difficulty)

Basics — see the status

Type into the chat box.

/context

Then a colored grid picture appears on screen. The more cells filled, the more capacity used, and it also shows as a number what percentage of the total you're using. You can see at a glance which items (system settings, files read, conversation content, etc.) take up how much.

Applied — expand the details

/context all

Append all and it expands to item-by-item details. Use it when you want to see concretely "which file is eating up capacity."

Advanced — find the cause and clean up

/context

First check the heavy item with /context (e.g., a big log file read in full), then clean up like below.

/compact

/compact is a command that summarizes the conversation so far to free up capacity. It keeps the essentials and trims the fluff, making room in the bowl again.

Common pitfall — /context only "checks," clearing is separate

Typing /context doesn't reduce capacity. /context is just a dashboard that shows the status. Once you've checked on the dashboard what made the car slow, the actual cleanup must be done with /compact (summarize/compress the conversation) or /clear (clear the conversation entirely). Use /compact if the topic continues, /clear if you're moving to a completely different task.

A real case

Claude Code once got noticeably slow mid-work. Turning on /context, I saw that one big log file read in full during debugging was taking up most of the capacity. I no longer needed that log, so I cleaned up the conversation with /compact and the speed came back. Instead of vaguely wondering "why is it slow," I pinpointed the cause and solved it.

Use it like this too

• See details: Expand to item-by-item details with /context all.
• Compress past 80%: Once capacity passes 80%, answer quality tends to drop, so clear with /compact before it fills up.
• Find the culprit: Check which big file is eating capacity, and build the habit of not reading it in full next time.
• Cleanup flow: See with /context → /compact if the topic continues, /clear if the topic changes.

Tip: When starting a new task, clearing once with /clear keeps the previous task's leftovers from mixing in, making answers more accurate.

How others use it

• Check "Context Rot" — where response quality drops as the conversation lengthens — with /context. — Claude Code Context Optimization (Infograb)
• Use /compact when continuity is needed, /clear when the topic changes completely. — Claude Code Session Commands (BSWEN)

Wrap-up

The key is one sentence. When it gets slow, check the cause with /context first. AI getting dumb often isn't a malfunction but a signal that its head is full. Check the cause with the dashboard (/context), and reclaim room with compression (/compact) or clearing (/clear).

Based on: Claude Code v2.1.154 (2026.05)