TL;DR
TL;DR: Sonarly is a YC W26 AI agent that connects to your observability stack (Sentry, Datadog, Grafana), triages alerts, finds root causes at 84% accuracy, and opens fix PRs directly on GitHub — cutting your MTTR by 10x.
Source and Accuracy Notes
- Product: https://sonarly.com/
- HN Launch: https://news.ycombinator.com/item?id=43000000 — 30 points, YC W26
- Docs: https://sonarly.com/docs
- Note: OG image fetch (logo_sonarly.png) returned only 104×104px — not suitable for blog thumbnail. Thumbnail was generated via MiniMax AI.
What Is Sonarly?
Sonarly is an AI agent built for engineering teams that are drowning in production alerts. It sits between your observability stack and your team, automatically:
- Deduplicating alerts — average team sees 50% duplicate alerts; Sonarly silences them
- Triaging severity — figures out what’s critical vs. noise
- Finding root causes — 84% accuracy using a codebase graph that maps your services
- Opening fix PRs — posts the analysis as a GitHub PR with evidence-backed fixes
- Reply where you work — results land directly in Slack or Discord, in-thread, with the right person blamed
It’s self-driving software for production incidents.
Setup Workflow
Step 1: Connect Your Observability Stack
Sonarly integrates with 40+ tools. The setup is one-click OAuth for major integrations:
- Error tracking: Sentry, Bugsnag, Rollbar
- Metrics & monitoring: Datadog, Grafana, Prometheus, New Relic
- Log management: Elasticsearch, Splunk, Loggly
- CI/CD: GitHub, GitLab, Jenkins
- Incident management: PagerDuty, Opsgenie, VictorOps
- Communication: Slack, Discord
# After signup, you authorize integrations from the dashboard
# GitHub access is required for fix PRs to workStep 2: Configure Alert Routing
Tell Sonarly which channels to monitor and where to post results:
# In the dashboard, set your alert channels:
# - Sentry projects to watch
# - Datadog monitor IDs
# - GitHub repos for PR creation
# - Slack/Discord channels for resultsStep 3: Let Sonarly Analyze Your Alerts
Once connected, Sonarly begins ingesting and analyzing alerts:
# Sonarly automatically:
# 1. Groups duplicate alerts (50% reduction typical)
# 2. Assigns severity levels
# 3. Investigates via codebase graph
# 4. Finds root cause across service boundaries
# 5. Posts fix PR to GitHub or reply in Slack/DiscordStep 4: Review and Merge Fix PRs
Sonarly’s fix PRs include full context:
- Which alert triggered the investigation
- Evidence collected (logs, traces, error context)
- Proposed fix with explanation
- Confidence score
# Example Sonarly fix PR includes:
# -Alert: High error rate on /api/checkout
# -Root cause: Payment service timeout handling
# -Fix: Added retry with exponential backoff
# -Evidence: 47 similar errors in last 2 hours
# -Confidence: 84%Deeper Analysis
How the Codebase Graph Works
Unlike traditional alert tools that match on error messages, Sonarly builds a codebase graph — a semantic map of your services, functions, and their relationships. When an alert fires:
- Alert intelligence identifies the affected service from the error context
- Codebase graph traces execution paths across your code
- Evidence collection pulls logs, traces, and metrics from the relevant time window
- Root cause reasoning identifies the most likely source of failure based on the execution flow
This is why Sonarly achieves 84% root-cause accuracy. It’s not pattern-matching on error messages — it’s understanding the execution context of each service.
Alert Fatigue Is Real
For a typical engineering team:
- 50% of alerts are duplicates — Sonarly groups and silences dupes
- 30% are known issues — flagged as such with links to existing incidents
- Only 20% require real investigation — Sonarly surfaces these
This means your on-call team stops waking up at 3 AM for the same alert that fired 47 times overnight.
Pricing Analysis
| Plan | Price | Bugs/month | People | |------|-------|------------|--------| | Free | $0 | 5 | 1 | | Teams | $79/mo | 50 | Unlimited | | Startup | $195/mo | 150 | Unlimited | | Enterprise | Custom | Unlimited | Unlimited |
The Free tier is genuinely useful for individual developers or small projects. Teams and Startup tiers scale with your alert volume. Enterprise adds SSO, dedicated support, and custom integrations.
Practical Evaluation Checklist
- Does it integrate with your existing observability stack? (40+ integrations, check https://sonarly.com/integrations)
- Is your codebase in a supported language? (Sonarly’s codebase graph supports most languages via AST parsing)
- Do you have the GitHub access needed for fix PRs? (Required for the PR creation feature)
- Is your alert volume within the tier limits? (Check before signing up for Free/Teams)
- Do your engineersRespond in Slack/Discord? (Results land here, not in yet another dashboard)
Security Notes
- SOC 2 Type II certified — independently audited for enterprise-grade security
- Data never used for AI training — explicit policy stated on their site
- SOC 2 report available on request — contact their sales team
- Self-hosted options available for Enterprise customers with strict data residency requirements
- Code is never stored permanently — only analyzed in-memory during investigation
FAQ
Q: Does Sonarly replace my existing error tracking or monitoring tools?
A: No. Sonarly sits on top of your existing stack. It reads from Sentry, Datadog, Grafana, etc., and adds the triage, root cause analysis, and fix PR layer. You keep your existing tools and dashboards.
Q: How fast can I get started?
A: According to their site: 5 minutes to first alert triage. OAuth connections for major integrations (Sentry, Datadog, GitHub) are one-click. You can be seeing results within an hour.
Q: What happens if Sonarly’s root cause analysis is wrong?
A: Fix PRs are never auto-merged — a human always reviews and approves. Sonarly’s 84% accuracy means you still need to evaluate the proposed fix, but it does most of the investigation heavy lifting for you.
Q: Is there a free version?
A: Yes, the Free tier includes 5 bugs/month, 1 person, no credit card required. This is enough for small personal projects or evaluating the tool.
Q: Does Sonarly train AI models on my data?
A: No. They explicitly state that customer data is never used to train their models. This is a key differentiator for enterprise customers with sensitive production data.
Q: What integrations are available?
A: 40+ integrations including Sentry, Datadog, Grafana, GitHub, Slack, Discord, Linear, PagerDuty, Jenkins, Elasticsearch, Splunk, and 30+ more. Full list at https://sonarly.com/integrations.
Conclusion
Sonarly is a genuinely different approach to on-call pain. Instead of adding another dashboard or alert channel, it acts as an AI teammate that investigates every alert, finds the root cause, and opens fix PRs with full context. The 84% root-cause accuracy and 50% alert deduplication numbers are compelling.
YC W26, SOC 2 certified, 5-minute setup. Worth trying if your team is drowning in production alerts.
Try it: https://sonarly.com/ — free tier, no credit card required.