Claude Code Unable to connect to API means the connection to the API never completed. Usually the network is fine and a leftover ANTHROPIC_BASE_URL, a dead proxy variable, a VPN…
llama.cpp shows "error: --model or -hf is required" when llama-cli starts without a model. Give it a local GGUF with -m or a Hugging Face repo with -hf. llama-server is…
Codex shows "Token exchange failed" when the request it sends by itself to auth.openai.com skips your proxy or never connects. Here is how to tell 403 from "error sending request"…
Claude Code settings.json location, checked against Anthropic docs on October 27, 2026: the user, project, local and managed file paths on each system, why the file may not exist yet,…
Six MCP servers picked by job: Playwright and Chrome DevTools for the browser, GitHub for PRs, Context7 for docs, Figma for designs and DBHub for databases. Each comes with its…
Codex CLI vs Codex App, checked against OpenAI docs on October 27, 2026: where each surface runs, what only the CLI or the desktop app can do, how the IDE…
How to point Claude Code at DeepSeek's Anthropic-compatible endpoint, what a heavy day costs at peak and off-peak rates, which Claude Code features stop working, and when Kimi or the…
A Claude Code 429 can be your Pro/Max session limit, your API key's rate limit or a monthly spend cap, and each needs a different fix. Exact messages, headers to…
A skill teaches Claude how to do something; MCP connects it to an outside service. What Anthropic's docs say about Claude Code skills vs MCP, hooks and plugins, which one…
Repeated 529 Overloaded errors in Claude Code mean Anthropic is out of capacity for that model. It doesn't count against your quota. How it differs from 429 and 500, what…
Ultrathink was reported dead in January 2026, came back in 2.1.68, and since Claude Code 2.1.311 it runs a terminal turn at xhigh. What each effort level does, whether the…