I was running a pool of Claude Code agents and realized every one of them held every credential I'd configured. scoped-mcp is what I built to fix that — and I built it using the same multi-agent pattern it now secures.
Using one coding agent is simple. Using several across multiple projects is where the real mess...
A technical deep-dive into architecting local-first, privacy-preserving AI agents using Go for orchestration and Rust for high-performance inference and security boundaries.
Claude Code in CI: Running Agentic Code Review, Test Generation, and Auto-Fix on Every Pull...