Wachd is a self-hosted alert intelligence platform that receives webhooks from the monitoring tools a team already runs, collects surrounding context, strips PII, correlates events into a causal timeline, and routes a plain-English diagnosis with suggested actions to the on-call engineer. On-call teams, SREs, and platform engineers use it to turn low-context pages into triaged incidents with on-call schedules, escalation, and Slack, email, voice, push, and Teams notifications.[1][2][3]
Quick verdict
On-call teams buried in low-context pages from Grafana, Datadog, or Prometheus benefit most from Wachd. Its decisive capability is PII-safe AI triage at alert time: it gathers commits, logs, and metrics, strips secrets and identifiers, then delivers a causal timeline with a probable cause and a suggested action to whoever is on call. The material constraints are youth and self-hosting weight: the repository dates to April 2026, several integrations are marked not yet tested, and production means operating Postgres, Redis, and the analysis backend yourself. If alert noise with privacy constraints is the problem and you can self-host, choose Wachd as the intelligent layer in front of Grafana; if you want zero-infrastructure paging, choose a managed responder instead.[1][2][3]
Key features
Inbound webhook intake Inbound webhook intake: Wachd receives alert webhooks from Grafana, Datadog, Splunk, Prometheus, and compatible sources over HTTPS, with no agents running in application clusters.[1]
Automatic context collection Automatic context collection: on each alert Wachd fetches recent commits, error logs, and metric history around the alert time from connected GitHub, Loki, and Prometheus datasources.[1]
PII sanitization before analysis PII sanitization before analysis: a sanitizer strips emails, IP addresses, account IDs, card patterns, API keys, hostnames, JWTs, and cloud resource identifiers before anything reaches the analysis backend.[1]
On-call schedules and escalation On-call schedules and escalation: Wachd looks up who is on call for the team, escalates, and supports per-user notification rules plus voice calls via Twilio.[1]
Who will love it
On-call teams with noisy monitoring On-call teams with noisy monitoring: teams receiving high volumes of low-context alerts from Grafana, Datadog, or Prometheus gain triaged diagnoses with probable cause and suggested actions.[1][2]
Privacy-constrained operators Privacy-constrained operators: organisations that need incident AI without sending data externally can run Ollama air-gapped with PII stripped before analysis.[1]
Who should avoid it
Young project with untested paths Young project with untested paths: the repository dates to April 2026 and GKE, Datadog webhooks, GPU Ollama, and Prometheus Alertmanager webhooks are marked not yet tested.[1][2]
Self-hosting operational weight Self-hosting operational weight: production means operating PostgreSQL, Redis, encryption keys, and the analysis backend, so teams wanting zero-infrastructure paging should prefer a managed responder.[1]
Alternatives and compatibility
Grafana complement Grafana complement: Wachd consumes Grafana alert webhooks and adds correlation plus on-call routing, so Grafana remains the monitoring layer; the complement edge against published Grafana is recorded in review.[1][2]
SSO and notification ecosystem SSO and notification ecosystem: OIDC sign-in via Entra, Okta, or Google Workspace with Slack, email, Twilio voice, APNs push, and Teams notifications downstream.[1]
Questions worth asking
Is Wachd really self-hosted?
Yes. Wachd is Apache-2.0 licensed and designed to be self-hosted with Docker Compose or Helm. The README states that data stays in your infrastructure and no agents run in your application clusters.[1]
Which monitoring tools send alerts to Wachd?
Wachd receives inbound alert webhooks from Grafana, Datadog, Splunk, Prometheus, and compatible sources, then enriches them with commits, logs, and metric history before analysis.[1]
Can Wachd run its AI analysis without external API calls?
Yes. Ollama is the default analysis backend for local, air-gapped deployments with no external API calls. Claude, OpenAI, and Gemini backends are also supported for teams that prefer them.[1]
Does Wachd replace Grafana?
Sources and editorial review
Source research. Reviewed on October 13, 2026. Written by Sanjib Sah. AI assisted; reviewed by a human editor.
We have not run or tested this product.
- Wachd homepage (official)
official · Observed October 13, 2026
Official homepage: self-hosted OpsGenie-replacement positioning, AI-assisted incident response, live demo, docs/GitHub/Discord links.
- wachd/wachd README (official)
official · Observed October 13, 2026
Repository README: inbound webhook flow from Grafana/Datadog/Splunk/Prometheus, Redis queue and worker architecture, PII sanitizer patterns, auth with OIDC SSO, Docker Compose and Helm deploy, tested deployment matrix.
- wachd/wachd LICENSE (official)
official · Observed October 13, 2026
LICENSE file carries the full Apache License 2.0 text; README badge and License section confirm Apache 2.0.
- wachd/wachd repository record
official · Observed October 13, 2026
wachd/wachd record: AI-powered alert intelligence platform, Go, created 2026-04-15; API license field unresolved so the LICENSE file governs.
- Grafana webhook notifier docs (official)
official · Observed October 13, 2026
Grafana Alerting documents a webhook contact-point notifier, the integration Wachd consumes inbound for alert intake.
Compared with these products
No explicit proprietary baseline is recorded.