Agentic Task System 0.2 — your task manager is the best agent memory you're not using

[renezander030]

TL;DR — ATS turns the task app you already live in (TickTick today; Notion / Obsidian / Things on the roadmap) into a first-class context layer for your coding agent. No new database, no separate "memory" store, no migration.

The thesis

Everyone's bolting a fresh vector DB onto their agent and calling it "memory." But you already maintain a curated, structured, mobile-editable record of what you're working on: your task app. ATS exposes it the way an agent queries — not the way you scroll it.

What you get

• Hybrid retrieval (RRF): dense + sparse + keyword fused via Reciprocal Rank Fusion, with sources: [...] provenance. The first fetch is the right one — fewer search → refine → search loops.
• Two-way bus: the same app you capture into from your phone becomes the agent's read/write channel (file attachments included).
• Capture-time relevance enrichment: each item gets hung on a "trunk" (an active project) at write time, so retrieval has structure to lean on.
• Adapter, not migration: keep your task app. The adapter interface is ~6 methods.

Install

npm install -g @reneza/ats-cli @reneza/ats-adapter-ticktick
ats auth login
ats find "the thing you half-remember writing down"

What changed in 0.2

Renamed from Agentic Knowledge Base (AKB) → Agentic Task System (ATS) to match the thesis: your task app is the context layer, not a separate knowledge store. Config/auth migrate automatically (no re-auth). Full notes: CHANGELOG.

Questions, adapter requests, and "here's how I wired it" reports are all welcome in this Discussions tab. 👇