a log explorer for the terminal · built for 2am

Pipe anything in. It already knows your fields.

lumen ingests raw JSON, logfmt, kubectl noise and stack traces, then gives you schema-aware search over your own keys. One binary. Zero config. Try cat prod.log | lumen.

$ curl -fsSL https://nightbuildlabs.github.io/lumen/install.sh | sh

or cargo install lumen-tui · binaries ↗

zsh — 96×28

this is the whole product ·

why it's different

The search bar reads your logs before you do.

/ schema autocomplete

Type three characters and lumen suggests keys from the data you just piped in — ranked by how often they occur. No schema files, no setup.

/ sta▍
  status_code   47,992
  status          3,120
  stage             214

▂▆█ error timeline

A sparkline along the border shows log volume over time, colored by severity. See the red spike, jump straight to it.

▂ 02:10
▃ 02:11
▄ 02:12
█ 02:13  ← jump here
▂ 02:14

| eats anything

Wrapped lines, interleaved streams, multiline stack traces, logfmt — the parser handles the mess. A line that can't be parsed is still shown and still searchable. Nothing is ever dropped.

$ kubectl logs api | lumen
$ docker compose logs | lumen
$ lumen dump-2026-07-04.log

an honest comparison

Plenty of good log tools exist. Here's where each wins.

tool	zero-config messy input	schema autocomplete	query power	learning curve
lumen	✓ core focus	✓	filters — SQL is roadmap	none — keys on screen
lnav	✓ known formats	—	full SQL — stronger today	steep
logana	✓ many sources	—	field / date filters	moderate
jless	JSON only	—	navigation only	low
tailspin	✓ colorizing	—	none — it's a highlighter	none
toolong	✓ + merges files	—	search only	low

Where another tool wins, we say so — lnav has stronger queries today. lumen optimizes for the first 60 seconds of an incident: pipe in, find the needle, copy it out. Full per-tool writeup, sources, and honest "use X instead when…" notes → docs/COMPARISON.md.

30 seconds from now

Your logs, already searchable.

$ curl -fsSL https://nightbuildlabs.github.io/lumen/install.sh | sh

macOS & Linux · single ~31MB binary (DuckDB included) · MIT licensed