Desktop client for Apache Kafka · Windows and corporate VDIs
Topicup opens clusters with thousands of topics in under a second, searches millions of messages without bloating memory, and runs where you actually work: Windows and corporate VDIs, with no installer and no admin rights. The conditions of every measurement
Open
One request for topics, partitions and brokers. A Rust engine built on librdkafka. A tree that only draws the rows you can see. This is how a real corporate cluster, read from the VDI, opens:
Metadata request for the whole cluster, including connection, TLS and SCRAM, no cache. Real corporate cluster with SASL_SSL, from the VDI, 16 October 2026, version 0.1 engine from the command line. The next two requests, on the open connection: 0.39 and 0.37 s.
Search
A tax code inside a JSON, across tens of millions of messages? Topicup tells you how long it will take before it starts, reads the whole topic, counts every match and keeps the first 1,000 for you to inspect, copy and export.
Test bench: 12 million JSON messages of about 500 bytes, 12 partitions, uncompressed, Redpanda in Docker on the same PC. Version 0.1 search engine, one run, 17 October 2026. RAM is the engine's: the window (WebView2) comes on top.
Safe
Topicup knows which environment you are in, and reminds you before anything that changes something.
Production connections, or those with no environment, write only after you unlock them. Development and staging write after a double confirmation.
In Windows Credential Manager. Never in a plain-text file.
Share your screen without showing customer data.
No telemetry: Topicup only talks to the clusters you point it at, never to us or anyone else.
Anywhere
Unzip it into a folder of your own, on the VDI or on a USB stick, and go. The connections you already have in Offset Explorer or KafkIO come over with an import.
Needs 64-bit Windows 10 or 11 with the WebView2 runtime, which is almost always there. Designed for VDIs: the engine has already read a real corporate cluster from a VDI; the full test of the window is in progress.
And everything else
KIP-848 groups too. Offset resets behind the safety catch.
Avro, Protobuf and JSON Schema, decoded into readable JSON.
New messages on top, with pause, resume and snapshot.
One message or many, with key, headers and partition.
Connectors, tasks, errors and queries, in the same window.
Who can do what, on which resource.
Your in-house formats, in sandboxed plugins.
Topics, connections and commands from one field.
Favourites, recents, groups and environments.
The same features, for your scripts.
Tested with Apache Kafka 4.3.1 in KRaft mode.
English and Italian, light and dark themes, keyboard and screen reader.
Side by side
12 s against 52 s for the runner-up and about 4-5 minutes for the others
Open the program, see the 10,003 topics, search for a value across 2 million messages.
Sum of the median startup, topic list and search. Kafbat UI did not finish the search.
4× faster than the runner-up, 20× or more than the others
A tax code across 2 million JSON messages, 972 MB: 200 expected matches.
Kafbat UI stopped after 5 minutes with 160,200 messages read. AKHQ and Redpanda Console: one run; AKHQ with a server setting changed to reach the end.
33 correctness questions, with Offset Explorer as the reference.
How we measure
A benchmark holds for its bench: network, cluster and data change the absolute times. That is why every number comes with where, when and with which version we measured it.
0.77 s
Real corporate cluster, SASL_SSL and SCRAM, from the VDI. First request, including connection, no cache. Version 0.1 engine, command line.
0.22 s
Local Redpanda in Docker, one broker. Median of 3 runs. Version 0.1.
2 min 2 s · 30 MB RAM
Local Redpanda in Docker, uncompressed data, one run. Engine RAM, without the window. Version 0.1.
Other clients' numbers are in the Side by side chapter and, with versions, licences used, method and raw data, on the comparison page. Method, licences and raw data
Blog
Want to try it?
We are preparing the first public version. No forms and no waiting lists: when the download is ready, you will find it here. If you want to try it sooner, on your own cluster, write to us.
Contact for early testing: [email protected]