Built for Celery & Redis · async queue monitoring

When your task queue
goes quiet, you'll know.

Your workers show as running. Your queue keeps growing. Nobody knows why. Kanari catches the silent failures generic monitoring misses — and tells you what to do about them.

Join the waitlist →or start free, right now
$ pip install kanari-agent

The silent failures

The incidents that never raise an exception

Your error tracker fires when a task throws. It stays silent when the worker dies mid-task, when the scheduler stops, when the queue quietly backs up. Those are the ones that page you at 3am.

👻

Ghost workers

A worker keeps its process up but stops consuming after a broker reconnect. Your dashboard shows it green. The queue tells the real story.

WORKER_OFFLINE

Invisible queue latency

200 tasks pending — but are they 2 seconds old or 20 minutes old? Depth alone lies. Kanari tracks how long the oldest task has really waited.

QUEUE_SLA_BREACH
🕳️

Silent task loss

A worker crash mid-execution with late-acks off means the task is gone — no retry, no error, no trace. You find out from the missing data.

CRITICAL config
🧱

Stuck tasks

One task running far past its expected runtime holds a worker slot hostage and starves everything behind it. Throughput collapses with no clear cause.

STUCK_TASK
📈

Growing backlogs

Tasks pile up faster than workers drain them. By the time you notice, you're thousands behind with no idea when you'll catch up.

QUEUE_BACKLOG

Capacity blindness

Workers look idle while queues grow. The real cause: prefetch hoarding tasks that can't run yet, blocking faster ones behind them.

HIGH_SATURATION

How it works

Diagnosis, not just another metric

Every finding comes with the probable cause, the command to confirm it, and a safe fix. Not "queue depth is 847" — but why it happened and what to do next.

🔍

Instant audit

One command, no config to start. Run it against any Celery + Redis setup in about 30 seconds.

🧠

AI-powered diagnosis

When an anomaly trips, Kanari turns raw signals into a plain-language explanation and a concrete next step.

Real latency tracking

Goes beyond depth to measure how long tasks actually wait, so SLA breaches surface before users feel them.

🔧

Config analysis

Catches risky settings like task_acks_late=False, weak eviction policies, and single points of failure.

⚙️

CI/CD ready

Clean exit codes and JSON output. Drop it into a pipeline as a pre- or post-deploy gate.

🔔

Alerts when it matters

Continuous monitoring that messages you the moment something breaks — and stays quiet when it doesn't.


Privacy-first by design

Your task data never leaves your infrastructure

The agent runs where your workers run. It reads health signals, not contents. Task arguments, payloads, and results are never accessed — not sampled, not hashed, never touched. What reaches Kanari is sanitized metrics, and nothing else.

✓ What leaves your infra
project: "acme-prod" heartbeat: "2026-06-29T03:14Z" workers: { alive: 3, total: 4 } worker_id: "w-a1b2" # hashed queues: [ { name: "emails", pending: 847, oldest_s: 125.4 } ] finding: "WORKER_OFFLINE"
✕ What never does
  • Task args and kwargs
  • Task payloads or return values
  • Message bodies from the broker
  • Connection strings or credentials
  • Raw worker hostnames (hashed first)
  • Anything that could identify your data or customers

Alerts

Routed into the tools your team already lives in

Kanari fits your on-call setup instead of asking you to watch one more dashboard. Start with Slack and email; wire in your full escalation stack as you grow.

SlackEmailPagerDutyOpsgenieWebhooksDiscord · soonMicrosoft Teams · soon

Slack & email come with Pro. PagerDuty, Opsgenie, webhooks and on-call escalation come with Team.


Your stack

Deep on Celery today. Broader tomorrow.

Kanari understands Celery's queue model, worker pool, and acknowledgment semantics natively — that depth is the whole point. The same engine is expanding to the rest of the async world.

Celery + Redis● Live

Full detection, config analysis, and diagnosis. The deepest support we offer.

RabbitMQ○ On the roadmap

Celery on RabbitMQ, plus native AMQP queue health.

Amazon SQS○ On the roadmap

Queue depth, age, and dead-letter visibility for SQS-backed workers.

Kafka○ On the roadmap

Consumer lag and partition health for streaming workloads.


Plans

Start free. Upgrade for peace of mind.

The agent is open source and free forever. The paid tiers add what needs a service running for you around the clock.

Free
The open-source agent. Run it on your own infra, forever.
Free · open source
  • audit & live watch
  • All detections built in
  • Deep config analysis
  • JSON output & CI exit codes
  • Privacy-first sanitization
View on GitHub
Pro
For the engineer who doesn't want to watch a terminal at 3am.
Pricing at launch
  • Everything in Free
  • Continuous monitoring, always on
  • Proactive alerts to Slack & email
  • AI diagnosis on every anomaly
  • History & trends
  • One project / environment
Join the waitlist
Team
For teams with an on-call rotation and more than one environment.
Pricing at launch
  • Everything in Pro
  • Multiple projects & environments
  • PagerDuty, Opsgenie & webhooks
  • On-call escalation
  • Per-project alert routing
Join the waitlist

Early access

Get on the list

We're opening Kanari to a small first group. Tell us where you are with Celery and we'll reach out as spots free up.

No spam. One email when your spot opens.