JJeevesy
Sign inBook a demo
AI-native · Agent-first · MCP throughout

The tenant operating system,
built AI-native from day one.

Jeevesy isn't a property platform with a chatbot glued on top. Every domain object, workflow and policy is designed as a tool an agent can reason over — with MCP as the protocol, LLMs as the runtime, and RLS-scoped identity as the boundary. The dashboard is a fallback for the humans who still like clicking.

How it's built Book a demo
Prefer buttons over prompts? Open the fallback dashboard.
Designed for agents, not adapted for them

A single MCP surface — the same interface the UI uses.

Cases, bookings, rent, documents, journeys and communications are exposed as typed, permissioned MCP tools. The web app calls them. Jeeves calls them. ChatGPT, Claude and Cursor call them. There is no second, "AI-only" API — because there is no non-AI path.

ArchitectureMCP endpoint
# Same tools, three surfaces
tenant_app →
create_case, list_bookings, list_payments…
jeeves_agent →
create_case, list_bookings, list_payments…
chatgpt/claude/cursor →
create_case, list_bookings, list_payments…
# Add Jeevesy as a remote MCP server:
https://jeevesy.lovable.app/mcp
Protocol-first architecture
MCP tools are the primary interface. The UI is one client among many.
Reasoning built into the workflow
Triage, drafting, and journey decisions are model calls, not if-statements.
Per-user identity end-to-end
Every agent call runs as the authenticated user with RLS enforced in the database.
Every assistant. Every surface. Every device.

MCP for any agent — plus deep native integration with the assistants people already use.

MCP is the open door for ChatGPT, Claude, Cursor and any future agent. On top of that, Jeevesy plugs directly into Apple Intelligence & the new Siri in iOS 27 via App Intents, and into Google Assistant & Gemini via App Actions and Connected Apps — so residents can raise a case, pay rent, book the padel court or ask about a leak without ever opening the app.

Apple Intelligence · iOS 27
Siri that actually does things in your building
  • • App Intents expose every Jeevesy tool to the new on-device Siri — voice, typed or from Spotlight.
  • • On-screen & personal context lets Siri reference the lease, unit or case you're currently looking at.
  • • Shortcuts & Automations for "when I get home", "every 1st of the month", or NFC tags at the front door.
  • • Live Activities & Dynamic Island for move-in day, package deliveries and open service cases.
  • • Home Screen & Lock Screen widgets, Control Center controls and StandBy for at-a-glance status.
  • • Apple Watch complications, CarPlay and Vision Pro spatial widgets.
  • • Foundation Models framework for private on-device reasoning where it matters.
Google Assistant · Gemini · Android
Say it to Gemini, done in Jeevesy
  • • App Actions & Gemini Connected Apps so Gemini can create cases, check rent or book amenities agentically.
  • • Gemini Spark agent-callable building blocks — every MCP tool mirrored as a first-class Android capability.
  • • Home Screen widgets, Material You theming and At A Glance on Pixel.
  • • Live Updates (Android 16+) for ongoing move-ins, deliveries and service visits.
  • • Wear OS Tiles & Complications, Android Auto, Quick Settings tiles.
  • • Google Home / Matter hooks for building-level smart devices.
Widgets everywhere
iOS, Android, macOS, Wear OS — the same tool schemas, rendered natively.
Live Activities & Live Updates
Real-time status for cases, deliveries and move-ins on the lock screen.
Watch & wearable
Apple Watch complications and Wear OS Tiles for one-tap actions.
CarPlay & Android Auto
Voice-first access on the way home — safe, hands-free, contextual.
Roadmap: Apple Intelligence & Siri (iOS 27), Gemini Spark App Actions, Wear OS Tiles and CarPlay/Android Auto integrations are on the near-term roadmap. MCP is available today.
Integration status

What's live today — and what's on the near-term roadmap.

We ship the assistant surfaces incrementally so you always know what production tenants can rely on right now versus what's in flight.

Available now
  • Remote MCP server
    Public /mcp endpoint with OAuth 2.1 — any MCP client (ChatGPT, Claude, Cursor) connects as the signed-in user.
  • Jeeves in-app agent
    Threaded chat console with auto-run MCP tool calls, confidence scoring and human handoff.
  • Governed memory
    Recall, remember, forget and tenant-initiated erasure — with hash-chained trace events.
  • Self-service tenant flows
    Create cases, book resources, cancel bookings and upload documents from chat.
  • Operator agent inbox
    AI summary, suggested reply, approve-and-resolve and escalation from a single thread.
  • Agent activity feed
    Every MCP call with reasoning summary, arguments, result and resulting state changes.
Planned
  • Apple Intelligence & Siri (iOS 27)
    App Intents for every MCP tool, plus on-screen context, Shortcuts and Live Activities.
  • Gemini Connected Apps & App Actions
    Agentic Gemini access on Android with Spark-callable building blocks.
  • Wear OS Tiles & Apple Watch complications
    One-tap case status, rent balance and booking check-in on the wrist.
  • CarPlay & Android Auto
    Voice-first, hands-free access to the concierge on the drive home.
  • WhatsApp & email concierge
    Same Jeeves persona and tool surface, delivered through the channels tenants already use.
  • Partner MCP directory
    Bring your own MCP servers (locksmiths, insurers, utilities) and expose them safely to Jeeves.
100%
of features available to agents
0
screen-scraping required
1
MCP endpoint, every surface
24/7
autonomous operation
The architecture

Intelligence in every layer — not just the chat box.

Most "AI" property tools bolt a chatbot on top of a legacy CRUD app. Jeevesy inverts that: the data model, the workflows and the interfaces are all designed to be driven by autonomous agents — with humans in the loop where it matters.

MCP-native domain

Every entity — case, booking, contract, invoice, journey — is an MCP tool with a typed schema and RLS policy.

Reasoning as a primitive

Case triage, category detection, sentiment, translation and drafting run through the LLM by default, not as opt-in add-ons.

Autonomous journeys

Move-in, move-out, renewals and reminders run as agent workflows with tool calls, retries and human-in-the-loop approvals.

Jeeves, in every surface

The same concierge lives in the tenant app, the operator dashboard, WhatsApp, email and any external MCP client.

Bring your own agents

Connect ChatGPT, Claude, Cursor, or an internal agent framework. Connect other systems as MCP servers back into Jeeves.

Tools > pages

Every dashboard page is generated from the same tool schemas the agents use. One source of truth for humans and machines.

Grounded communication

Notices, replies and translations are retrieved and grounded in property knowledge — no hallucinated policies.

Insight loops

Adoption, deflection, sentiment and SLA metrics feed back into prompts and journey policies automatically.

Safety by construction

Mutating tools require approval, every action is audited, and secrets never enter the model context.

Legacy property tech
  • • CRUD app with a chatbot bolted on
  • • AI is a sidebar; the real work happens in forms
  • • Integrations via nightly CSVs and screen scrapers
  • • "AI" means autocomplete and a summarize button
Jeevesy
  • • Agents are the primary interface; the UI is a client
  • • Reasoning is a first-class layer of the stack
  • • MCP in and out — for tenants, operators and partners
  • • Autonomous journeys with human-in-the-loop where it matters

See what agent-native property tech feels like.

Book a 30-minute walkthrough, or connect Jeevesy to your own agent in under a minute.

Book a demoTry the fallback dashboard