AI incident investigation documentation

Availability: available

Mendry is a self-hosted AI incident investigation system for teams that need to connect production signals with logs and deployed code. It supports an incident-response workflow from evidence collection and root cause analysis to a reviewable repair proposal, while merge and deployment decisions remain with people.

Start with the source-evaluation workflow below. Check product status for supported capabilities, installation readiness, and work still in progress.

Investigate your first production issue

Follow the getting-started workflow to prepare the environment and configure a project.
Connect a supported monitoring signal through signed webhooks or review Tencent CLS.
Inspect the first incident, its evidence, diagnosis, and proposed next steps.

Explore current features

Use the feature map to find console modules, configuration boundaries, and detailed guides.
Follow the operator workflow for day-to-day investigation and remediation review.
Check product status before relying on a preview capability or evaluating a production rollout.

Understand the architecture

The Agent Harness is the underlying execution architecture. Its extension contracts support development of the investigation system; they are not the product’s primary operator journey. The shared foundation is implemented and focused-tested, while full migration of the incident loop remains staged work.

Read architecture, lifecycle, and security boundaries.
Read Agent Harness and Extend the Harness for execution and extension contracts.
Check Local walkthrough status for the separate development path. Its prerequisites and readiness differ from the incident application.

Documentation policy

Commands appear only when current code, tests, or a tested release artifact supports them. The Local source tree is not treated as a validated onboarding release, and the production installation page remains blocked until its artifact and support boundary are verified.