Yes. The self-hosted edition is free and open-source under the MIT license, with no per-GB or per-host fees.

Do my logs leave my server?

No. Detection runs entirely on your host. Optional AI root-cause analysis sends only anomaly summaries, never raw logs, and is opt-in.

A self-hosted, open-source AI log analyzer and monitoring tool that flags the few log lines signalling a real failure and explains them in plain language.

Engine v0.13.0 · MIT · 0 lines uploaded

The AI log analyzer that never ships your logs anywhere.

LogLens AI reads logs where they already live, flags the few lines that signal a real failure, groups repeats into incident families and explains each one in plain language. No baselines, no labelling, no training. Install, point, read.

Self-hosted · zero-touch · MIT open source · 0 log lines uploaded

Try it in your browser

Copied · +10 XP. Next: loglens analyze --source app.log

1,400+PyPI downloads

450+Docker Hub pulls

~7×faster warm runs

0lines uploaded

MEASUREDWarm daemon ≈2.4 s → ≈0.33 s. Downloads as of Oct 2026.

loglens watch · kubectl logs -f checkout
{{ lensHint }}

{{ l.ts }}
{{ l.lvl }}
{{ l.msg }}

score {{ l.score }} · why: {{ l.why }}

Illustrative stream
{{ anomalyCount }} anomalies flagged

PLAYGROUND · NO INSTALL

Feel it in 30 seconds. Pick a mission.

Run real LogLens commands against sample incidents. Press / to focus the terminal.

REPLAY MODE · recorded v0.13.0 run

{{ t.text }}

operator@lens$

try: loglens --help · echo $? · whoami

Incidents
{{ canvasMeta }}

{{ f.lvl }} · {{ f.svc }} · ×{{ f.count }}
score {{ f.score }}

{{ f.msg }}

why: {{ f.why }}

Run this on your own servers. Install Get Premium early

INCIDENT FAMILIES

{{ collCount }} errors → 1 incident.

Your engineers see one incident with a count, a score and a reason, not a wall of red. Keep scrolling.

INCIDENT · db×{{ collCount }} · score 0.95

connection refused | host=db-1

why: severity ERROR, failure keyword, burst

WHY SELF-HOSTED

Built for the people accountable for uptime.

0 B

Data stays sovereign

Logs are read, parsed and analyzed on your machines. AI root-cause is opt-in and sends anomaly summaries, never full logs.

₹0 / GB

Cost stays predictable

It runs on hardware you already operate, so log growth never turns into a surprise invoice.

1 binary

Operations stay simple

Installed from your normal package manager. Registers as a systemd, launchd or Windows service. Stable exit codes.

▶

1:16 · no signup {{ ovHint }}

SEE IT WORKING

Watch before you install.

The why in 76 seconds. Hover to preview. The tour pauses at key moments so you can run each command yourself.

24LogLens reads the logs
48Your logs never leave
00Secure by design

Watch the interactive tour Walkthrough · 1:57

HOW IT WORKS

Six stages. All on your host.

DAY ONE

What your team gets, free, today.

ROADMAP

From log analyzer to AI monitoring tool.

Logs are the first and strongest signal, so that's where we started. Application monitoring with metrics and traces is in the pipeline.

● LIVE · v0.13.0LogsAnomaly detection, incident families, CI gate, offline reports.

IN THE PIPELINE · Nov–Dec 2026MetricsPrometheus and OTLP metrics.

IN THE PIPELINE · Q1 2027Traces + change eventsOTLP traces and deploy events for "what changed?" root cause.

IN THE PIPELINECluster modePremium Go engine. Many nodes' cores scheduled as one machine, with failover.

INSTALL

Deploys like any other system package.

Native installers bundle neural mode and need no Python. New in v0.13.0: one-line installers, Windows MSI, Scoop and Intel macOS builds.

{{ i.cmd }}

Gates your pipelines

Stable exit codes, JSON output and the official GitHub Action, now with sticky PR comments, SARIF upload and report artifacts.

0Success, no gate tripped

1Error: bad source, model or input

2--fail-on gate tripped: fail the build

- uses: LoglensAI/LogLens-AI@v0.13.0
  with:
    source: "build.log"
    fail-on: error
    comment-on-pr: true
    sarif: true

LOGLENS ACADEMY

Docs you can play. Rookie to Chief of Uptime.

Start Level 1 · 4 min

PATH {{ p.n }} · {{ p.levels }} levels {{ p.name }} {{ p.rank }} {{ p.desc }}

PLANS

Core is free and live today.

Pro, Business, Apex and Enterprise run on the new Apex Go engine with AI monitoring. They're in the pipeline, and prices are revealed at launch.

Explore every plan →

LIVE · FREE · MIT Core ₹0 · $0 Open-source engine, 1 node, unmetered. Install now IN THE PIPELINE Pro Revealed at launch Apex Go engine on every core of one box. Tap to see all features IN THE PIPELINE Business Revealed at launch Up to 5 nodes, HA leader, RCA graph, metrics. Tap to see all features IN THE PIPELINE Apex Revealed at launch 16 nodes, AI narration, auto-heal, traces. Tap to see all features IN THE PIPELINE Enterprise Custom Unlimited nodes, SSO + SCIM, SLA. Talk to us

SECURITY & DATA FLOW

Know exactly where your data goes.

Component	Where it runs	What leaves the host
Parsing & detection	On the host, as the invoking user	Nothing
Warm daemon	Unix socket (0600, per user) or loopback TCP with a random token on Windows	Nothing, localhost only
Neural model (--deep)	Bundled with native installs; local directory via LOGLENS_MODEL_DIR	Nothing, runs offline
AI root-cause & ask	Only when you pass --rca or use ask	Opt-in. Anomaly summaries to your chosen provider, with count and token usage reported.
Alerts	Only when you configure an alerter in the SDK	Opt-in. The alert message, to the Slack, Teams or email destination you set.

QUESTIONS FROM IT LEADERS

The short answers.

WAITLIST · DESIGN PARTNERS

Planning a rollout? Talk to the person who builds it.

Your message goes straight to Paras Rajput. Design partners get free Apex during Preview.

01Share your setup: servers, log sources, compliance needs.

02Get a personal reply by email.

03Hear first when Premium opens.

● RECEIVED Thank you. You'll hear back by email.