NeuBird Features · Investigations
Root cause, with the work shown.
Investigations lists every investigation NeuBird has run. Each one shows the signal that triggered it, the root cause analysis and the fix, and every finding carries a citation back to the exact query the agent ran and the log line it found, so you can check the work instead of trusting it.
Last 24 hours| ID | Investigation | Status |
|---|---|---|
| inv-2291 | Checkout latency: orders-db pool exhausted | Fix ready |
| inv-2290 | flink-flight-processor TargetDown | Resolved |
| inv-2288 | Payments 5xx after config change | Resolved |
| inv-2287 | auth-svc token refresh errors | Running |
kubectl rollout history deploy/checkout-api -n prod --revision=42Evidence it foundREVISION 42 deploy #4821 replicas: 6 → 12For illustration only. The product interface may differ.
How Investigations works
What is in an investigation?
Open an investigation from the table and it shows three things in order. The signal is what triggered it. The root cause analysis is the chain of events that caused the problem, delivered in under 5 minutes at 94% accuracy. The fix is the change that resolves it, with a Merge PR button to apply it.
How do citations work?
Every claim in a root cause carries a small numbered citation. Opening one shows where the agent got that piece of evidence: the source it queried, the exact command or query syntax it ran, and the log line or metric it pulled.
An engineer can rerun the same query and see the same line. That makes each finding something you can validate, not a conclusion you have to take on faith.
Can I see the agent's reasoning?
Yes. As an administrator, you can view the agent's full thought process for any investigation, from the first query to the conclusion, alongside every source it drew on. Every agentic action is audited.
Who approves the fix?
Your engineers do. The fix arrives as a pull request, and nothing merges until an engineer clicks Merge PR. Every action is human-approved, and the approval is recorded with the investigation.
Each concluded investigation is recorded as operations memory: conclusions, causal chains and approvals, with zero telemetry storage. The next time the same failure appears, the agent starts from what your team already learned.
FAQ
Common questions
How fast does NeuBird find the root cause?
NeuBird delivers root cause analysis in under 5 minutes at 94% accuracy, analyzing 15+ sources in parallel. Each investigation records the triggering signal, the root cause and a proposed fix.
How can I verify an investigation's findings?
Every finding in a NeuBird investigation is cited. A citation shows the source the agent queried, the exact command or query it ran, and the log line or metric it found, so an engineer can rerun the query and confirm the evidence. Administrators can also view the agent's full reasoning.
Does NeuBird apply fixes automatically?
No fix runs without human approval. NeuBird proposes the fix as a pull request, and an engineer merges it. Autonomy is set per environment as Suggest, Recommend or Act, and every action is human-approved and audited.
NeuBird Investigations
See it on your own stack.
Book a demo and watch NeuBird work across your environment, with every action approved by your team.