APACHE-2.0 · SELF-HOSTED · NO VENDOR LOCK-IN

Open-Source Alerting for Your Existing Stack

Keep your metrics and logs where they are. Write and run alert rules across Prometheus, VictoriaMetrics, Elasticsearch, ClickHouse, and more — from one place. Send alerts to Slack, PagerDuty, or whatever your team already uses.

Built-in AI assistant
MCP server

Try live demoNo sign-up

Install in 5 minutesSelf-hosted

GitHub★ 13k+

Run the whole stack locally

git clone https://github.com/ccfos/nightingale.gitcd nightingale/docker/compose-bridgedocker compose up -d

S1CPU usage above 90%

__name__=cpu_usage_idleident=n9e-demo-01busigroup=DevOps

Fired 02:14:07Firing for 14m

Delivered

13,327GitHub stars

10+Data source types

20Notification channels

86Built-in integrations

74MCP tools

HOW IT WORKS

From any data source to someone getting paged

Nightingale does not take over your storage. It does one layer well. Here is how data moves through it — each stage is a real object in the product.

datasourceruleeventagent

event — data becomes an alert here

datasource

Connect what you already run

No migration. Register Prometheus, VictoriaMetrics, ElasticSearch, ClickHouse or Loki as a data source and Nightingale evaluates alerts on top of it. Nothing collecting yet? Categraf covers operating systems, middleware and databases in a single agent and writes straight in over Prometheus Remote Write.

10+ data source types — metrics, logs and databases alike
86 built-in integrations bundle collector config, dashboards and rules

rule

Write the rule once

Metrics, logs and SQL-queried stores share one rule model, so there is nothing new to learn per data source. Rules are configured in the UI, scoped by business group, and graded S1 / S2 / S3. A library of common rules ships with the install, so you never start from an empty page.

One rule can span several data sources
Edit rules in the UI or through the API — no YAML reload cycle

event

Cut the noise, then route

Storms collapse into an aggregated view by business group and severity. Mute rules, subscriptions and event pipelines filter what is left, and the survivors reach people through 20 notification channels. Wire up ibex when an alert should fix itself.

DingTalk, Feishu, WeCom, phone, SMS, Slack and 15 more
Event pipelines group similar alerts, rewrite labels and route conditionally

agent

Hand it to an agent

Since v9 the n9e process is itself an MCP server. Claude Code, Cursor and anything else that speaks MCP connect to /mcp and manage alerting in plain language. There is no extra process to deploy.

Nightingale AI assistant and Skills included
Permissions follow the token — identical RBAC to that user in the UI

~/.claude.json

{
  "mcpServers": {
    "nightingale": {
      "type": "http",
      "url": "http://127.0.0.1:17000/mcp",
      "headers": { "X-User-Token": "<your-token>" }
    }
  }
}

BUILT-IN MCP SERVER

74 tools, nothing extra to deploy

The MCP endpoint lives inside the n9e process, so it is not one more component to operate. Every tool call is dispatched onto Nightingale’s own HTTP API carrying your token — a client can never reach anything its token’s owner cannot.

42read tools

32write tools

13toolsets

Toolsets

alertstargetsdatasourcemutesbusi_groupsnotify_rulesalert_subscribesevent_pipelinesusersmetricslogsdashboardsroles

The write tools are not registered until you turn them on in config. Authenticate with a personal token or OAuth 2.1: Nightingale can act as the authorization server itself, so hosted clients like Claude and ChatGPT connect with zero pre-registration, or as a resource server in front of your existing Keycloak, Entra ID or Okta.

which S1 alerts from last night are still firing?

nightingale · get_alert_events severity=1 status=firing

3 still firing. Longest: mysql-prod replica lag > 30s, 2h14m

what did memory on mysql-prod look like this past hour?

nightingale · query_metrics ident=mysql-prod range=1h

Peak 87.4%, now 62.1%. No memory rule fired.

mute the replica lag alert until 9am, reason "change window"

nightingale · create_mute_rule scope=mysql-prod until=09:00

✓ Mute rule #1024 created, expires tomorrow 09:00

INTEGRATIONS

It grows on the stack you have

Nightingale never asks you to replace what works. Data sources, collectors and notification channels all plug into what is already running.

DATA SOURCES · Data sources

Prometheus
VictoriaMetrics
ElasticSearch
ClickHouse
Loki
VictoriaLogs
MySQL
PostgreSQL
TDengine
OpenSearch
Doris
IoTDB

NOTIFICATIONS · Notification channels

Slack
PagerDuty
Email
Telegram
Discord
Jira
Mattermost
Phone / SMS
FlashDuty
DingTalk
Feishu
WeCom

See every integration →

DEPLOY

Run it where you need it

Self-hosted

Apache-2.0, fully featured. One compose command brings it up; binaries and Helm are there too.

docker compose up -dDownload and install →

Edge data centers

Deploy n9e-edge as a local alerting engine where the link to the center is unreliable. Alerts keep firing and keep being delivered through a network partition.

n9e-edge --configs etc/edgeAbout the edge architecture →

COMPARISONS