Agentic Quartz
An AI assistant for individual real-estate agents that works across the tools where a deal already lives.
Per-cycle cost cut from ~$0.10 to under $0.01 with a custom multi-tier memory system. In private beta with working agents.
Problem
Individual real-estate agents — not brokerages with an ops team — juggle dozens of contacts, follow-ups, and scheduling tasks across email, calendar, files, and phone. They lose deals to missed follow-ups and forgotten context. Existing CRMs and brokerage platforms answer this by asking the agent to move their whole day onto a new dashboard, which means manual data entry that solo agents won't do consistently.
Approach
Built an assistant that fits around the tools agents already use instead of asking them to move into a new dashboard. It connects to Gmail, Calendar, and Drive read-only to start, and the whole product is organized around an explicit permission model — observe, then draft, then act — where each level is enforced in code and the user's own settings, not promised in a prompt. Today it observes and drafts; it does not send mail, send texts, or make edits on its own until the agent turns that on, level by level. Under the hood, the system observes Gmail via Pub/Sub push notifications, with a prefilter service that blocks promotional noise before the agent even wakes. A 3-tier memory system gives the agent persistent context tailored to real estate: a curated set of facts loaded into every prompt, an archive of observations searchable via semantic embeddings, and a structured contact graph with full interaction history. The agent graduates its own knowledge between tiers, keeping a pressure valve that forces it to curate rather than accumulate. Calendar events are cross-referenced with known contacts for the morning briefing. A 4-layer cost optimization stack (pre-check skipping, light context for quiet cycles, daily session resets, cheaper models for routine work) compounds to roughly 90% savings on idle heartbeats. Each user runs in an isolated rootless Podman container with their own SQLite database, OAuth credentials, and notification channel. The infrastructure layer is being extracted into HiveClaw, a product-agnostic Go runtime targeting sub-50MB memory per instance, designed to scale to 50+ persistent agents on a single 8GB host.
Result
After optimization, per-cycle cost dropped from $0.10 to under $0.01. Agentic Quartz is in private beta, built alongside practicing residential agents against their real daily workflow. It triages the inbox, keeps each deal's context together across email, calendar, and files, and delivers a short morning briefing — with zero manual CRM entry — while staying read-only and draft-only until an agent explicitly widens its permissions. The infrastructure layer is being generalized into a framework that replaces 1.5GB-per-container runtimes with a sub-50MB Go process.