Running AI agents in production
Running AI coding agents in production — notes from a practitioner. Claude Code, Codex, opencode, MCP servers, LLM gateways, and the boring ops that keep agents useful.
Claude Code
Codex
opencode
MCP
LLM gateways
Latest posts
- A repeatable build pipeline: lint, test, and gates in one command2026-08-16
- About2026-08-16
- API key hygiene for LLM calls: headers, proxies, and kill switches2026-08-16
- Background processes in automation: nohup is dead, track them properly2026-08-16
- Backup discipline: what I back up, when, and how I prove it works2026-08-16
- Bridging MCP Servers with mcporter: Ad-Hoc Tools Without Config Changes2026-08-16
- Checking Domain Availability with RDAP: No WHOIS, No Rate Limits2026-08-16
- Choosing a Registrar: First-Year Prices Are a Trap2026-08-16
- Choosing an embedding model for Japanese text2026-08-16
- Closing port 22: SSH over Tailscale only2026-08-16