What is Microtica Agents

Microtica Agents gives your team an AI investigator for infrastructure incidents. You ask what’s wrong — or mention it under an alert — and the Investigator digs through your connected accounts and comes back with a case: a working hypothesis, the evidence behind it, and the next moves. It’s built for small dev teams who run real infrastructure but don’t have a dedicated SRE.

The console mid-session: the alert that started it, the Investigator's root-cause answer, and the case board with conclusion, risk, and confidence A session in the console — what it found, and the case board beside it.

The three surfaces

Chat. The Investigator works in your Slack or Discord the way a teammate would. Mention @Microtica under a Slack alert, or run /investigate in a Discord channel, and it digs through the connected accounts and reports back in the same thread with what it found. When it needs something from you, it asks — in the thread, not in a dashboard you forgot to open.

The inbox. Every case also lives in the console, in one list ordered by whether it needs a human. What’s waiting on you sits on top and reads larger — a question, a failed run, a watch that tripped. Everything quiet sits below it in one dim line each. Watches the Investigator is holding for you collapse into a single heartbeat at the foot of the list. You never hunt through chat history to find out what’s still open.

Cases. An investigation is a case, not a chat log. Each case shows a Working hypothesis with a plain-language confidence, Findings individually marked Confirmed, Likely, or Low, and concrete Next moves. Follow-ups don’t start over — they add a chapter to the same case’s journal, so the history of an incident reads front-to-back in one place.

What the Investigator can reach

Once you connect an environment, the Investigator gathers evidence across it:

AWS — ECS, EC2, RDS, Lambda, CloudWatch and more, through a read-only role that lives in your own account.
Kubernetes — pods, deployments, events, and logs, through read-only access you apply yourself.
Git providers — your repositories: code, commits, diffs, branches, and pull requests, to connect a symptom to the change that caused it.
Cloudflare — DNS, Workers, and edge configuration, through an API token you scope and paste.
MCP servers — any remote MCP server you connect, adding its tools to the Investigator’s built-ins.

You decide what it can see, and every grant is inspectable before you make it — the AWS template and the Kubernetes manifest are both open for review, and access can be revoked at any time. The details are on Security & access.

Connecting your first environment takes about two minutes, and your first case usually lands a few minutes after that.

Next: Quickstart