CloudThinker Resolve
Finds the why. In minutes.
Resolve investigates every incident across your whole stack at once, finds the root cause and proposes the fix. Every incident makes it smarter.
Most teams still wake an engineer at 2am.
Usually for problems an AI should handle. And detection without action is just a louder pager.
Detect, Analyze, Resolve, Validate.
Resolve finds the why, proposes the fix and proves it worked. Then it remembers, so the next incident goes faster.
Detecting 13K events / 60s
14 sources
- 08:43:025xx rate 18% on /checkout
- 08:43:01payments-db-prod CPU 97%
- 08:42:58checkout-svc p99 latency 3.1s
- 08:42:55orders-worker retries climbing
- 08:42:49Paged on-call
Correlated
12 signals became one incident, routed to the database investigation.
F88
Root cause in minutes. Not days.
“We started with a small read-only pilot, verified the results, and expanded from there. Now tasks that used to take days or weeks can finish in minutes or run automatically.”
Infrastructure Manager, F88
Read the storyYour team stays in control.
Resolve proposes. Your engineers decide what it may do on its own.
You approve the fix
In Manual, each tool is set to always allow, needs approval or never allow. A fix that changes something waits for an engineer.
Auto Mode and runbook rules
In Auto, every write is checked before it runs. Each runbook command is allowed, waits for approval, or is denied.
Every action on the record
Who did what, when, with what privilege and what outcome. The runbook and post-mortem are written when the incident closes.
Your policy
Runs, waits for approval, or stops.
Every action
Logged and reversible.
Your access
Brokered credentials. Sandboxed execution.
Questions, answered.
What is CloudThinker Resolve?
Resolve is CloudThinker's Frontier Investigation Agent for incidents: a team of specialist AI agents that detect, analyze, resolve and validate production incidents. It clusters raw signal into incidents, investigates the root cause in parallel, and remembers every incident so the next one starts smarter than the last.
How is Resolve different from AIOps?
AIOps platforms compress operational signal (log analysis, metric correlation, anomaly detection) and surface incidents for a human to investigate. Resolve takes the compressed signal as input and runs the investigation, picks the runbook, executes the response inside a sandbox with scoped credentials, and writes a tamper-evident audit record. AIOps stops at the alert. Resolve carries the action through to a reversible, approved production change.
How does Resolve cut alert noise?
Resolve clusters the signal first. It ingests alerts from observability tools (Datadog, Prometheus, Grafana, Splunk, ELK, PagerDuty, Opsgenie), de-duplicates and correlates them into a single incident, and starts parallel investigations. The team stops paging on noise.
How does Memory make incidents faster over time?
Every resolved incident (root cause, steps taken, outcome) lands in CloudThinker's Memory store. The next time a similar pattern appears, Resolve retrieves the prior resolution, avoids previously failed approaches, and surfaces recurring patterns for permanent engineering fixes. MTTR drops over time.
How does Resolve perform on public incident benchmarks?
On the 21-fault SREGym-Lite cohort, a live Kubernetes incident benchmark, Resolve with GPT-6 Astra diagnosed 100% of faults and mitigated and resolved 100% of the 20 faults whose mitigation check can pass. One fault has a check that no model can pass, so we void it. The best public leaderboard row resolves 81.0%. With Claude Opus 5, Resolve resolved 90.0%. Both runs were self-run, one cold run per fault, and are not official leaderboard submissions. The full seven-model evaluation is published on the CloudThinker blog.
Engineers get their time back. For creative work.
Agents investigate every issue across application, cloud, security, database and Kubernetes, find the why and propose the fix. Your engineers approve, under your policies and audit trail.