dailynotes
Personal intelligence agent that ingests daily signals (meetings, Slack, Jira, etc.) and builds a queryable knowledge base with nightly briefings and real-time context for Claude.
21 commits over 4 days
What it is
Daily Intelligence Agent consolidates fragmented professional and personal signals—meeting transcripts, Slack messages, Jira assignments, calendar events—into a single self-hosted knowledge base. It runs automated nightly pipelines, processes high-priority events via webhooks in near real-time, and delivers concise morning briefings. The system also exposes an MCP server so Claude can query accumulated context (tasks, contacts, meetings, relationship profiles) during every conversation, turning months of data into an always-available intelligence layer.
Why I built it
Built to solve the problem of lost context and missed follow-throughs in a high-velocity professional life. The developer needed a way to own all their data, compound intelligence over time, and reduce cognitive overhead by surfacing the right information at the right time. The PRD explicitly frames this as a tool for managing high-context workloads while maintaining full data ownership (no third-party subscription for the knowledge store).
Tech stack
Next.js 16 (App Router), TypeScript, React, Tailwind CSS v4, shadcn/ui, Drizzle ORM, Neon (Postgres with pgvector), Trigger.dev (scheduled tasks), React Email + Resend (briefings), OpenAI embeddings, Anthropic Claude (LLM classification & synthesis), NextAuth, Sentry, Vercel (deployment), GitHub Actions CI/CD with preview databases per PR.
Timeline & effort
Last active Mar 19, 2026 — 21 commits across 4 active days (spanning 27 days).
Active development over 27 calendar days (Feb 20 – Mar 19, 2026) with 21 commits concentrated on 4 distinct days. Commit clustering suggests intense sprint-style work sessions rather than daily iteration—likely a fast-paced build driven by a clear PRD and tight scope for a solo user.
Did it turn out useful?
This is a shipped, functioning tool with production deployment (Vercel), live webhooks (Krisp, Slack, Jira), a comprehensive dashboard, nightly automation via Trigger.dev and Vercel cron, and a working MCP server for Claude integration. The PRD is at v2.7, CI/CD is mature (lint, test, preview DBs, Sentry), and the setup wizard and multi-phase backlog show thoughtful productionization. It solves a real, personal problem for the developer and is actively used (deployment URL is live). While it's a single-user tool, the engineering is robust and the problem it solves is ongoing.