Internal operations platform Ongoing

Suiki Agent

An autonomous operations agent that watches production signals, investigates, and prepares proposals and draft pull requests. Draft creation can be automatic within configured limits; changes remain available for human review.

Suiki Agent service architecture with Sentry, GitHub, Cloudflare, Grok and Slack logos
Service architecture overview ยท investigation, drafts, human review and Slack notifications
Team AI-augmented lead with a security advisor
Configured policies Safety model
Continuous Scheduler
Full Audit trail
Policy-controlled External actions

The Challenge

Production support means repetitive first-response work: reading error signals, checking failures, gathering context, and deciding whether something needs action. The goal was to make that loop dependable without giving an AI agent unsafe write access.

What We Built

An operations agent that watches production signals, classifies them, enriches investigations with safe context, and prepares proposals and draft pull requests. Draft creation can be automatic within configured limits, with evidence available for human review.

Configured Slack notifications and operations summaries keep the owner informed. GitHub remains the review surface for proposals and draft pull requests.

Safety & Evidence

Task state, results, and operational status are recorded for review. Approval requirements depend on the action and configured policy. Draft creation can be automatic; proposals and draft pull requests are available for human review.