Raindrop Workshop debuts local agent debugger

This is a strong “debugging layer” bet, not just another observability dashboard. The interesting part is the workflow: local traces, replay, eval generation, and fix verification all chained together so agent development gets closer to a real feedback loop.

• –Local-first tracing lowers the friction for iterating on agents without shipping logs to a SaaS first
• –MCP support is the sharpest angle here, because it lets a second agent inspect and act on traces directly
• –The self-healing loop is compelling, but only if teams actually trust the replay and eval outputs enough to automate decisions
• –Open source plus free makes this easy to trial, which matters in a category where adoption usually dies on setup pain
• –It sits near the overlap of debugging, evals, and observability, but the product story is really about closing the loop on agent failures