Guides for running coding agents
Short, specific answers to the problems that come up when Claude Code, Codex or OpenCode is your daily driver.
Claude Code hit a rate limit. Now what?
What a Claude Code 429 or usage limit means, which errors are worth retrying elsewhere, and how a fallback chain keeps the session going.
Read moreYour prompt cache is the biggest line on the bill
Cache reads cost about a tenth of normal input. When the cached prefix changes, you pay the full write price again. How to see it and what a proxy must not do.
Read moreKeep /compact working on any model
Why /compact can fail when Claude Code runs on another provider's model, and how a gateway keeps the summary turn text-only.
Read moreStop long reasoning streams from dropping
Why long-thinking model responses drop mid-stream behind proxies and CDNs, and how protocol-correct keepalives fix it.
Read moreWeb search that doesn't depend on the model
Give agents web search through one MCP tool backed by your own search provider, whichever model is answering.
Read moreOpenCode vs Claude Code
OpenCode and Claude Code compared for daily coding, and how to run both on the same models, fallback chains and usage view.
Read moreCode search and memory for every agent
Semantic code search, documentation indexing and cross-session memory for Claude Code, Codex and OpenCode from one MCP server.
Read more