Introducing logcat.ai Playground

Board bring-up, an OS port, a transient bug that shows up once every thousand boots. Something fails and what lands on your desk is a pile. A kernel log from one tool, a trace from another, a packet capture somebody grabbed at the same time, an archive from the vendor with four hundred files in it. The failure is in there. It’s also spread across artifacts nobody designed to reference each other, each one needing its own viewer, and finding it manually costs days.

logcat.ai reads that pile and tells you what broke.

Today we’re opening Playground to anyone with a file and an email address, free. No sales call, no credit card. Bring the log that’s costing you the most time right now.

Every capability is included#

Playground gives you the whole platform. Quick Search, Deep Research and Delta comparison run at full analysis depth. Nothing has been cut down for the free tier.

Quick Search answers in under five seconds when you only need to know whether a subsystem is throwing warnings. Deep Research runs a multi-step investigation across your files together and connects events between subsystems, like tying a driver crash in the kernel log to a CAN bus dropout three seconds earlier, and looks up CVEs and external references while it works. Delta takes two or more captures in any supported format and isolates what changed between versions, devices or builds.

Everything comes back cited to the log line it came from. Findings export as PDF, and any analysis can be shared as a public link with the citations intact.

What you can upload#

Linux: dmesg, syslog and journalctl exports across ARM, x86, MIPS and RISC-V, from distribution servers to embedded boards
Android: bugreport zips including OEM-nested ones, logcat, ANR traces, tombstones, dumpsys
Automotive: AAOS bugreports, VHAL dumps, CAN bus traces in candump format
Telecom: decoded exports from QXDM and QCAT, MediaTek ELT, Shannon DM
Network: pcap and pcapng, plus btsnoop Bluetooth HCI logs
Binary formats: Perfetto traces and SQLite snapshots, decoded into queryable surfaces
Bundles: a zip, 7z or tar of mixed logs is unpacked, analyzed file by file, then correlated across the set
Anything else: unknown text formats are classified at ingest, so a log format only your team uses still gets read

The complete list is in supported file types.

Limits#

250 MB of uploads per account, total
enough for about five deep investigations a month

The 250 MB is cumulative across everything you upload. Deleting a file doesn’t return the space. There’s no per-file cap, so a single large bundle can use most of it. Both caps will rise as our capacity allows.

Different operations draw down different amounts, and a comparison across more files draws down more than a single-file look. How the budget works is in investigation units and quotas.

Playground has no top-up. When that stops being enough, or if you need an isolated environment for proprietary code, write to founders@logcat.ai and we’ll set your team up on a dedicated deployment.

What happens to files you upload is covered in the terms.

Put it to the test#

Bring a capture you already know the answer to, the bug that took your team three days last month. That’s the test we’d run in your position.

If it finds your bug, tell us. If it misses, tell us that and what the root cause turned out to be. We can’t read every format yet, so name the one we’re missing. Most of what Playground handles today is there because an engineer asked for it.