Errors and fixes

zsh: command not found: claude - the actual fix Find where it actually installed, why the shell cannot see it, and the fix that survives a restart. 7 min Claude Code rate limit reached: what to do right now Which limit you hit, what to do in the next five minutes, and how to stop it recurring. 14 min Claude Code not working: a diagnostic order that finds it Six checks in the order that finds the problem fastest, plus a symptom-to-cause table. 8 min Claude Code permission errors: reading the refusal correctly Filesystem denials, policy refusals, macOS folder protection, and the sandbox all look alike. How to tell which you have. 8 min Claude Code context low: what it means and what to do What the warning means, why compacting deliberately beats waiting for it, and how to fill the window slower. 9 min MCP server not connecting in Claude Code The stdout trap that breaks most custom servers, plus the six other causes in order. 6 min Claude Code in WSL: the problems and their fixes The filesystem trap that halves your speed, searches that silently miss files, and login that never returns. 6 min Codex CLI not working: the sandbox is usually why Most Codex failures are the sandbox working correctly. How to tell those from real problems, starting with one command. 11 min Delete a Claude Code session, and everything it left behind Why /clear deletes nothing, where the transcripts live, and the command that removes a project cleanly. 7 min Exit code 137 and the rest: reading how a process died What 137, 143, 127, 126, 130, 124, 1, 2, and 255 actually tell you, and what to do about each. 9 min Claude rate limits: two systems that share one name A subscription window and an API throughput limit are different mechanisms with the same name. Here is both, and how to tell them apart. 7 min Claude rate limit reached: is it you, or is Claude down? Five different messages, three of them about you and two about Anthropic. Which one you got decides everything. 6 min Claude token limit per day: the ceilings that actually exist Nobody publishes a daily token cap because there is not one. Here is every ceiling that does exist, in its real unit. 7 min Unable to connect to Anthropic services: diagnose it in order Whether it is Anthropic, your network, your proxy, or your credential, and how to tell in sixty seconds. 14 min Cursor says "You've hit your usage limit": which pool fired The exact checks to identify the empty pool, keep working safely, and prevent the same mid-task surprise next cycle. 18 min Why Cursor is slow, and how to isolate the real cause Typing lag, slow Agent replies, high CPU, and Remote SSH stalls have different causes. This sequence separates them in minutes. 12 min API error: 529 from Anthropic: what overloaded_error means and how to survive it A capacity signal, not a quota signal. Retry it properly, then fail over, and stop reinstalling things. 9 min API error: 400 messages: text content blocks must be non-empty The whole 400 family in one place, with the exact string each condition produces and where agent frameworks generate them. 10 min You exceeded your current quota, please check your plan and billing details A 429 that no amount of retrying will clear, and four different causes now hide behind it. 9 min Incorrect API key provided: the OpenAI 401 and every reason you get it Five different 401s, one message everybody quotes, and an environment variable you forgot you exported. 8 min This model's maximum context length is: fixing context_length_exceeded What actually counts toward the window, the real limits in August 2026, and the four ways out ranked by what they cost you. 9 min The server had an error while processing your request: OpenAI 500, 502, 503 and 504 Two of these four come from OpenAI. The other two usually come from something between you and OpenAI, and the fix is different. 8 min RESOURCE_EXHAUSTED on the Gemini API: which of the two 429s you have Two different 429s, a rewritten error format, and a limits table Google quietly deleted. What is actually knowable. 9 min The model does not exist or you do not have access to it Four different problems wearing one error message, and a rename table for August 2026 so you can tell which is yours. 10 min LLM API timeout errors: which layer actually killed your request Your SDK waits ten minutes. Cloudflare gives up at 125 seconds and your load balancer at 60. Whoever is strictest wins. 9 min