Why Does Claude Keep Crashing? Fixes and Work Recovery
Claude crashes, hangs, or fails to connect for several predictable reasons. This guide covers the most common causes, the quickest fixes, and how to save your work brief before restarting so you do not lose progress when a session ends unexpectedly.
Updated 5 min read
Claude crashes, hangs, or fails to respond for a few predictable reasons: a service outage, a rate limit, a context window that has grown too large, a browser issue, or a network problem. Most of these are recoverable in under five minutes once you know which one you are dealing with.
The harder problem is losing the work brief when a session ends. If the context was all in the chat, a crash means starting from scratch. Tools like fenbs keep the project brief, current tasks, and decisions outside the conversation, so restarting or switching to a different assistant does not mean rebuilding the context.
Check if there is a service outage first
Before troubleshooting your browser or network, check whether Claude itself is down. Anthropic maintains a status page at status.anthropic.com. If there is an active incident, wait for it to resolve — no amount of refreshing or restarting will fix a server-side problem.
If you need to know quickly whether a reported issue is widespread, see is Claude down for how to check current status and what to do while waiting.
Common causes and fixes
If the status page shows no incident, one of the following is likely causing the problem:
| Symptom | Likely cause | Fix |
|---|---|---|
| Spinning or no response | Server load or temporary outage | Wait 30 seconds and retry. Check status.anthropic.com. |
| Conversation stops mid-reply | Context window approaching its limit | Start a new conversation. Paste only the essential context. |
| App freezes on iOS or Android | Cached state or memory pressure | Force-close the app and reopen it. Update to the latest version. |
| Page not loading in browser | Browser cache or extension conflict | Try a private or incognito window. Disable extensions one at a time. |
| Rate limit error message | Too many requests in a short period | Wait a few minutes before retrying. Reduce request frequency. |
| Connection error or timeout | Network instability or VPN interference | Check your internet connection. Try disabling the VPN temporarily. |
The context window problem
A long conversation fills Claude's context window. When a conversation approaches or exceeds the limit, responses slow down, stop mid-sentence, or fail entirely. This is a fundamental limit — the model can only process a certain amount of text at once.
The fix is to start a new conversation and bring only the essential context. This means knowing what that context is before the session crashes. For ongoing work, the essential context is the project goal, current decisions, and the immediate next step — not the entire conversation history.
Saving work before a session ends
The single most useful habit for crash recovery is saving progress before a session gets too long. Do not wait for the crash; ask Claude to summarize the current state of the work periodically during a long session. Specifically:
- The decisions made in the session, with the reasoning.
- The current state of each task: done, in progress, blocked.
- The immediate next step for each open task.
- Any information that should not be re-derived in the next session.
Paste that summary into a project tool before closing the session. When you restart — whether by choice or after a crash — you paste that summary into the new conversation and continue.
Save progress before the session ends
Ask a Claude session connected to fenbs to save the current task states and update Where we left off before you close the conversation. If the session crashes, the next one reads that record and continues from the right point without rebuilding the context from memory.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
Claude Code keeps crashing
Claude Code, the agentic coding tool, can stop for different reasons than the Claude.ai web app. Common causes include: a long context filled with file reads, a tool call that times out, a permission error on a file, or an issue with the project's AGENTS.md or configuration file. Check the terminal output for the specific error. A context too long for the model is often the root cause of unpredictable stops.
For Claude Code specifically, keeping a clear AGENTS.md with the current task state helps a restarted session continue from the right point. See what is Claude Code for more on how Claude Code manages context across sessions.
How to do this with fenbs
- Create a free account. Add a project for the ongoing work where crashes have caused the most disruption.
- Connect Claude to fenbs from Integrations. At the start of each session, ask it to read the project context so it starts with the current brief.
- During a long session, ask Claude to save task progress and update Where we left off in fenbs every hour or before switching topics.
- If the session crashes, open a new conversation, ask it to read the fenbs project context, and continue from Where we left off.
- After a crash, note the approximate context length and the task that was running. If a specific task type causes repeated crashes, break it into smaller steps.
When crashing is a sign of a deeper problem
If Claude crashes consistently at a specific point in a project — always when a certain file is pasted, always after a certain number of messages — that pattern points to a specific cause. Investigate it before the next session rather than continuing to rebuild context after each crash. Common fixes include splitting a large input into smaller parts, starting a fresh conversation at the beginning of each major task, or storing the working context in a project tool so the chat itself stays shorter.