A case, not a chat log
Each investigation opens a case with a working hypothesis, a confidence level and the next moves. Follow-up questions land in the same case.
Deep InvestigationBeta
Find the root cause of AWS incidents. Approve the fix.
Ask Mica what broke. It reads AWS, Kubernetes, logs, deploys and config, then returns the root cause, the evidence and a fix you approve.
Read-only by default. Nothing changes until you approve it. A separate app from Microtica, in beta.
- Demo project
- Mica
- checkout-api restarts
Sample Mica conversation. Asked why checkout-api keeps restarting in Production, Mica checks the environment's health, the service and its logs, and compares versions across environments. It finds that the pods are killed for exceeding their 256 Mi memory limit since commit 9f3c1a2 added an in-memory product cache, shows the service and its logs, and asks for approval to raise the memory limit to 512 Mi in Production.
How it works
Ask, investigate, get a case.
- Step 1
Ask
Describe the problem in plain language. No query syntax and no jumping between dashboards.
- Step 2
Investigate
Mica plans the checks, runs them read-only and connects metrics with logs, events and recent deploys.
- Step 3
Get a case
The root cause, the evidence behind it and a recommended fix. Any change waits for your approval.
What you get
Built to investigate, not to chat.
Mica works through your connected accounts the way an engineer would, and tells you which parts it hasn’t pinned down.
Findings graded by evidence
Each finding is marked Confirmed, Likely or Low. Dead ends stay visible as Ruled out, so you can see what the answer rests on.
A journal across runs
Every run adds a chapter with its verdict and full report. A long incident reads as one story.
Watches
Ask Mica to keep an eye on something after a fix. It checks on a schedule and closes the watch only once it verifies the fix.
Memory and skills
Mica remembers facts about your setup and loads your runbooks as skills, so you don’t re-explain your infrastructure.
Triage or full depth
Start with a quick triage. Choose Deepen when you need the loose ends chased down.
Where you ask
Start where your team already talks.
Cases started in Slack post back to the thread, so the team sees each decision where the conversation is.
- Console
- Ask Mica at app.microtica.ai and follow the case as it fills in.
- Slack
- Mention @Microtica in a thread. The case posts back to the channel.
- Discord
- Run /investigate in a channel and /continue in its thread.
- MCP clients
- Give Claude Code or your own MCP client an API key to start and follow investigations.
Use cases
Whatever paged you, start by asking.
Cost spikes, crash-looping pods, slow endpoints, broken deploys. The same plain-language question, and a case back every time.
Cost spikes
“Our AWS spend is up 30% this week. Why?”
You get
The resource or deploy that changed, with the cost trail.
Crash-looping pods
“The api-server pod in prod keeps restarting.”
You get
Out of memory or exit code, the deploy behind it and the fix.
Fargate task failures
“My ECS service won’t stay healthy.”
You get
Task metrics, health check settings, IAM and the cause.
Latency regressions
“The API got slow around 2pm. What changed?”
You get
Metrics, logs and events lined up on one time range.
Deploy-induced incidents
“Did the last deploy break checkout?”
You get
The change matched to the error spike.
Permissions and connectivity
“The service can’t pull from ECR.”
You get
The IAM policy or security group rule to fix.
Access and approvals
It can look. It asks before it touches.
You decide which accounts, clusters and namespaces Mica can see, using credentials you create and revoke on your side.
How security works- AWS
- A CloudFormation stack you launch creates an IAM role with the AWS-managed ReadOnlyAccess policy. Every call shows up in your CloudTrail.
- Kubernetes
- You apply a service account with get, list and watch permissions. No verbs that change the cluster.
- Approvals
- When a step needs approval, the run pauses until someone approves it. Proposed fixes use a separate permission you grant, never the read-only role.
- Scope
- An alert-triggered investigation stays in one AWS account. Crossing into another connected account needs a person to approve it.
- Revocation
- Delete the stack or the service account and access ends.
- Self-hosting
- You can host the app in your own cloud account, so evidence, credentials and cases stay in your environment.
Works with the setup you have
Mica connects to your AWS accounts and Kubernetes clusters directly. Your services don’t have to be deployed with Microtica.
Read the docsTry the beta
Run your first investigation on the free plan. For paid plans, talk to us.
See pricingAsk about your last incident.
Connect an AWS account in a few minutes and see what Mica finds.