Your cloud incidents,
investigated automatically.
OpenDevOps connects to your cloud logs and metrics — AWS, Azure, and beyond — analyzes anomalies with AI, and delivers root cause analysis directly to your Slack. No more context switching at 3am.
03:41:15 UTC — payment-service Latency p99 > 5s
03:41:20 UTC — API Gateway 504 Timeout rate increased
An on-call engineer that never sleeps
Comprehensive analysis built for scale and security.
Autonomous detection
Automatically picks up alerts from PagerDuty or DataDog and starts investigating immediately.
Multi-Cloud
Native integrations across AWS and Azure — CloudWatch, CloudTrail, Azure Monitor, and more. Zero setup required for basic metrics.
Root cause in minutes
Reduces MTTR by summarizing complex log correlations into plain english explanations.
Any LLM
Bring your own API key. Supports GPT-4, Claude 3, or run Llama 3 locally for full data privacy.
Read-only by design
Read-only access policies ensure OpenDevOps can only inspect metrics and logs, never modify your infrastructure.
Delivered where you work
Investigations are posted directly to Slack or MS Teams incident channels as threaded replies.
Real incidents. Real cloud. Scored against ground truth.
10 reproducible AWS + Azure scenarios. The agent investigates over the same API a user hits — no cherry-picked transcripts. Run it yourself with make eval.
For engineers. A faster wrong answer is worthless on call. 90% on a commodity model — frontier models score higher.
For SRE leads. The agent correlates alarms, logs, metrics and CloudTrail in ~1 min — work that takes an engineer tens of minutes by hand.
For buyers. The marginal cost is the LLM token bill — a few cents on commodity models, dropping as open models improve. No per-seat or per-second platform tax.
Cost comparison — 10 incidents, same wall-clock time
| Method | Total time | Total cost (10) | Per incident |
|---|---|---|---|
| Manual triage (20–40 min @ ~$100/hr) | — | ~$33–67 | ~$50 |
| Managed cloud agent (~$0.008/s assumed) | 541.6 s | ~$4.33 | ~$0.43 |
| OpenDevOps (gpt-oss-120b via OpenRouter) | 541.6 s | ~$0.25–0.40 | ~$0.03 |
Competitor rate is assumed (~$0.008/s) — verify against published pricing before quoting externally. Cost figures use OpenRouter dashboard actuals. Model: gpt-oss-120b, run 2026-06-02.
Start free. Scale when you're ready.
Open Source
The full agent on your own infra. Forever free.
- checkFull AI observability & agent testing
- checkTrue OSS: same codebase as hosted
- checkAWS + Azure integrations
- checkAny LLM via your API key
- checkSQLite / Postgres storage
- checkSelf-hosted, full control
- checkYou control data retention
- checkCommunity support
Free Cloud
Try the managed service. No credit card required.
- check10 team members
- check30-day data retention
- check20 investigations / month
- checkAWS + Azure integrations
- checkCommunity support
Pro
For growing engineering teams that need more.
- checkUp to 50 team members
- check60-day data retention
- check500 investigations / month
- checkAll 30+ cloud integrations
- checkSlack & MS Teams notifications
- checkEmail support
Enterprise
For organizations with compliance and scale requirements.
- checkUnlimited team members
- checkCustom usage plans
- checkEverything in Pro
- checkVPC / private deployment
- checkSSO / SAML
- checkSOC2 & HIPAA compliance
- checkCustom data retention
- checkDedicated success manager
Frequently asked questions
Anything missing? Open a discussion on GitHub.
How do I self-host OpenDevOps?expand_more
Self-hosting OpenDevOps is straightforward — download the code from our GitHub repo and follow the README to get started. The agent is up in minutes.
What is the difference between the plans?expand_more
The main differences are in usage limits and data retention. All plans include the full suite of cloud investigation and AI agent features. Enterprise plans additionally offer custom hosting and deployment options, advanced identity and authentication management, personalized support, increased regulatory compliance, and more.
Which cloud providers are supported?expand_more
AWS and Azure are supported today. GCP support is coming soon.
Which LLMs does it work with?expand_more
Any model accessible via OpenRouter, Anthropic, or OpenAI — configure it in Settings → Agent config. Frontier models (Claude Sonnet/Opus, Gemini 2.5 Pro) give the highest accuracy. Mid-tier open models like gpt-oss-120b reach ~90% accuracy at ~$0.03 per investigation. You are never locked to one vendor.
Does it have write access to my infrastructure?expand_more
No — read-only by design. The agent queries logs, metrics, and config to build its investigation. It never modifies, restarts, or deletes any resource. On the managed plans you connect your AWS account via a cross-account IAM role scoped to read-only actions.
What happens when I hit my monthly investigation limit?expand_more
On Free Cloud, new investigations are paused until the next billing cycle resets. On Pro, contact us to top up or upgrade. Your existing session history remains fully accessible regardless of your limit.
How is my data stored and retained?expand_more
Investigation sessions and results are stored in Postgres. Free Cloud retains data for 30 days, Pro for 60 days, Enterprise with custom retention agreed at contract time. On the self-hosted plan you control storage entirely — SQLite or your own Postgres.
Is there a free trial for Pro?expand_more
Yes — Pro comes with a 14-day free trial. A credit card is required to start the trial. You get the full feature set and will not be charged until the trial period ends.