copperbox is a small workshop of open-source tools built for people who'd rather ship than babysit. Health checks that don't lie, dashboards that don't sprawl, plugins that earn their keep — made because we needed them, shared because you might too.
An observability service for incident response. Point your OTLP exporters at it and it learns your entire service topology from the traces you already emit, then draws it as a live, pannable map — so you spot an incident at its source instead of bisecting it team by team.
An MCP server that hands AI agents a real knowledge base: plain Markdown files with frontmatter, indexed into a link graph and served over MCP for search, traversal, and authoring. No database, no embeddings — and it opens as an Obsidian vault too.
copperbox started as a place to keep proactive-deps — a small Node library for service health checks — and slowly turned into a whole shelf of tools for the same problem: knowing which of your services are actually healthy, and why the unhealthy ones got that way.
Our taste: do one thing, ship a single binary or a tiny npm package, write a README that fits on a screen, and treat observability as a feature for humans, not a SaaS line item. If a tool needs a webpage to explain it, the tool is wrong. (Yes, we are aware of the irony of this webpage.)
Have a service that won't admit it's dying? Try tracemap. Its older sibling depsera ballooned out of scope and started doing too much; tracemap was started from those learnings — much more focused, with the lowest friction to onboarding. Point your OTLP exporters at it and you're done.