Kafka
Produce to and consume from a Kafka cluster, in the same window as everything else.
A Kafka tab connects to a cluster, sends messages to a topic, and tails one topic while you watch. It speaks the same broker protocol the Java and librdkafka clients do, over the same brokers and the same wire, so what works here works from your services.
The layout
A connection bar runs across the top; the work happens in two panels below it.
Producer
Value · Key · Headers · Schema · Config, then Produce.
Consumer
Subscribe, then read, filter and export what arrives.
One tab is one connection. Producing and consuming share it, so a message you send is one the same tab can be watching arrive.
Three things worth knowing up front
Watching takes nothing away
The consumer assigns partitions directly. It never joins a consumer group and never commits an offset, so tailing a topic here does not rebalance the applications already reading it, and does not move anyone else's committed position. You can watch production traffic without touching it.
The history lives in the backend
What you read is held in Rust, under a byte budget, and the panel shows a window onto it. A busy topic fills that buffer, not the webview, so the search runs over everything held, and the UI stays responsive while thousands of messages go by.
A tab lasts as long as the app is open
A Kafka tab is not saved anywhere. The address, the security settings, the message you were composing and everything the consumer has read live in memory only, and closing the tab or the app takes them with it. There is no Save here and nothing to reopen tomorrow, so a cluster you use often is worth putting in an environment variable rather than retyping. The same fact is what makes the security fields safe to type into: nothing you put in them is written down.
What a tab can do
Every security shape
SASL (incl. OAUTHBEARER), TLS, and mutual TLS.
Manage topics
Create, describe, edit and delete a topic without a CLI.
Full producer control
Key, partition, headers, acks, compression, tombstones.
Start anywhere
Earliest, latest, an offset, a time, or the last N.
Read, filter, export
JSON, text, binary or Avro; search it all; write JSON, JSONL or CSV.
Schema Registry
Encode a produce against the registry's schema, decode what you read, or bring your own Avro schema.