January 28, 2026
AI config cage match
Show HN: Config manager for Claude Code (and others) – rules, MCPs, permissions
Tool promises to wrangle your AI helpers; fans cheer, purists cry bloat
TLDR: Coder Config centralizes AI assistant settings, adds memory and project groups so tools remember context. Commenters split between loving the simplicity and warning it’s “config-of-config” bloat with privacy risks, making it a lightning rod for teams juggling multiple repos and seeking fewer setup headaches.
Show HN dropped “Coder Config,” a tool to tame AI coding chaos, and the comments turned into group therapy. Fans swooned over Workstreams linking related projects, a single registry for those add‑on “MCP” tools (extra tools you hook to the AI), and persistent memory so Claude remembers “use our logger.” One dev called it “adult supervision for my repos,” and loved the multi‑tool output across Claude, Gemini, and more. The local Web UI on port 3333 got love: clicky toggles, no more JSON spelunking. Fans also liked that plugins can be turned on per project, so shared tools are everywhere while sensitive ones stay scoped for teams.
Then the skeptics arrived. Minimalists blasted it as “config for your config,” insisting dotfiles and scripts are enough. Security folks worried that memory across sessions could store secrets or bleed context between projects, turning convenience into compliance drama. Others side‑eyed the rename from @regression‑io/claude‑config to coder‑config, reading it as hedging away from Claude—defenders said the alias still works. Jokes flew: “microservices support group,” “my AI now remembers my bad decisions,” plus the classic “works on my machine.” Salvation or bloat, the thread agreed on one thing: config is pain—and ripe for drama.
Key Points
- •Coder Config is a configuration manager for AI coding tools, supporting Claude Code, Gemini CLI, Codex CLI, and Antigravity.
- •The tool introduces Workstreams, a Unified MCP Registry with global→workspace→project inheritance, and hierarchical rules.
- •Persistent Memory stores preferences and patterns across sessions, and a Plugin System supports LSP servers, MCP tools, and custom commands.
- •Multi-Tool Output generates configuration files for different assistants, enabling switching without reconfiguration.
- •A local Web UI runs on port 3333; installation is via npm (Node.js 18+), with CLI commands for projects, MCPs, memory, environment, and workstreams.