A personal AI assistant that reads every inbox, triages by importance, and fixes its own bugs.
Monitors three email accounts, Telegram, WhatsApp, and iMessage. Sends SMS alerts for P8+ items only. Delivers a morning digest with a calendar-aware action list. Auto-fixes CI failures and resolves Sentry issues by opening pull requests.
The problem
Owner-operators don’t have inbox problems — they have signal problems. There’s no shortage of messages. There’s a shortage of time to figure out which three matter before noon.
What DK Echo does
- Monitors Gmail, Telegram, WhatsApp, and iMessage in one pipeline.
- Classifies every message by priority (P1–P10) using Claude.
- Fires SMS alerts for anything P8+ — and only P8+.
- Delivers a 6am morning digest with context-aware summaries, today’s calendar, and a resolution check (so threads you already replied to don’t show up as “action required”).
- Detects CI failures (GitHub Actions, Vercel, Sentry) and auto-fixes them: clone → Claude fix → pull request → auto-merge on green → resolve the Sentry issue.
- Has a dashboard, a chat assistant, and a morning briefing email that works on mobile.
The architecture that made it cheap
The hard part of agentic systems is usually the AI bill. DK Echo runs on the Claude Code CLI under a
team subscription — the same tool developers use for coding, repurposed as an inference engine
via claude -p. Marginal cost per classification: zero.
Every piece is boring, proven tech: Node.js, SQLite, Express, launchd, Cloudflare Tunnel. Nothing novel. The novelty is in the integration and the prompt engineering.
Why it’s in this portfolio
DK Echo is the template. The same pattern — inbox monitoring, priority routing, background agents, auto-fix loops — applies to customer support triage, operations alerting, sales handoffs, and finance anomaly detection. If your team runs on email, there’s a DK Echo shape for your business.