One afternoon a message never arrived. I chased it through the app's logs, a Postgres table and a Kafka topic, in three windows, and lost the thread every time I switched. In the end it was sitting in the topic, not consumed.

To see what my local stack was doing, I had five tools open: Docker Desktop, a database client, a Kafka UI, k9s and a terminal. Each had its own window and login, and none of them knew about the others. Every part of the stack lived in a different tool, and the code lived in none of them. If you run a stack locally, you may have the same five windows open right now.

So I built one panel in VS Code, next to Terminal and Ports, and called it Services. Now I open the project and the whole stack is there: no logins, no five windows to arrange.

Docker's live charts, then a Postgres connection, a table filtered with WHERE and the console, all in the Services panel

Private, fast, secure and free

  • Private: no telemetry, no account, no cloud; it talks only to your Docker Engine, your kubeconfig's clusters and the connections you save.
  • Fast: nothing connects until you expand it, and each client loads only the first time you use it.
  • Secure: passwords stay in your OS keychain, and AI assistants get only the connections you allow, through a local MCP server that's off until you turn it on, each connection set to off, read, or read and write.
  • Free: because I think developer tools should be.

Install it

One tree for the whole stack, with nothing to set up

What runs in Docker or Kubernetes shows up in one tree: Compose projects and their containers, pods, Deployments and Services. Select a container and its logs follow beside the tree; open a shell in it; watch its CPU, memory against its limit, network and disk as live charts.

A container's live charts, another container's logs followed, then a shell in it, in the Services panel

The part I wanted most: the databases and brokers inside those containers show up as connections without any setup.

  • A Postgres container is a Postgres connection, with its databases, tables and a console that completes your tables and columns.
  • A Kafka container is a Kafka connection, with its topics, consumer lag, and a topic you can follow like tail -f.
  • A Redis container shows its keys grouped by prefix.

Cassandra, MongoDB, Elasticsearch and S3 show up the same way.

So the message I was chasing would have been one tree away: the app's log, the row that never changed, and the topic's consumer lag showing nobody had read it. No second login, no copying a port into another tool.

What's next

Next is editing table rows. Today you can read, filter and export them as CSV or JSON. It's in early development, so expect rough edges; your issues decide what comes after.

If it doesn't work on your machine

Every issue comes straight to me: it always works on my machine, and I'll make sure it works on yours too.

A keycap says "Works on my machine…", three keycaps frown, then smile when it adds "…and it will work on yours too."

Who's behind it

I'm Marko, from Serbia. I build the tools I wished I had in my own work, the Bitno extensions. They're free, because I think developer tools should be.

Marko Milic on LinkedIn: Agentic Engineer, Java Backend and Developer Tooling, creator of Bitno VS Code Extensions

References