For devs with more than one desk

Your AI setup
shouldn't reset every room.

You code at home, refine at the office, ship from a laptop—and every switch means re-explaining context to your agent. LEMdesk keeps one local model, one knowledge base, and one thread across your LAN. Starter is the free sample. Pro is the desk you actually run.

See the demo Try Starter free Join Pro waitlist
  • 3 roomsone thread
  • 100%on your LAN
  • $0Starter forever
Python Cursor MCP OpenAI-compatible LAN-only 
$ git clone github.com/billrilea-lab/LemDesk.git
$ python3 lemdesk_auto.py --fast --skip-dmr-check
 
✓ knowledge corpus refreshed
✓ RAG index rebuilt
✓ handoff.md — edit by hand
✓ MCP server ready for Cursor
 
→ Sample kit ready. Try it free.

The problem

Every room is a cold start.

Cursor doesn't sync chat history across machines. Cloud AI doesn't know your local docs. And nobody wants to re-configure models, MCP servers, and skills every time they change desks.

Lost context

You spent an hour in a thread at home. At the office, the agent has no idea what you were building.

Repeated setup

Model URL, MCP config, skills, RAG corpus—copy-pasted or rebuilt from memory every switch.

Cloud lock-in

SaaS desks charge per seat and send your prompts elsewhere. Local models deserve local continuity.

LEMdesk's job: wire your LAN so model, knowledge, and agent context travel with you—not through a cloud, through files and endpoints you own.

Why two tiers

Starter proves it.
Pro makes it effortless.

Open source gets you in the door—one afternoon to see local desk sync work. Pro is what converts curiosity into a daily habit: health at a glance, handoff in one click, sync without opening a terminal.

Starter Sample

  • CLI sync + RAG corpus
  • MCP doc search in Cursor
  • Manual session handoff .md
  • Supervisor scripts
Try Starter free

Pro Full power

  • Menu bar — live desk health
  • Smart Handoff — paste-ready agent prompt
  • Dashboard — DMR, RAG, MCP at a glance
  • One-click sync from the menu bar
Join Pro waitlist

The stack

Six pieces. One desk.

Starter ships all of this free. Each piece solves one part of the multi-room problem—together they replace the re-setup ritual.

One model URL

Why: Stop installing models on every machine.

Point every desktop at one OpenAI-compatible endpoint on your LAN—:12434/engines/v1.

Auto knowledge

Why: Your docs should follow you, not live in one laptop.

Refresh a local RAG corpus in seconds. MCP tools in Cursor query your markdown—not a vendor cloud.

Session handoff

Why: Cursor won't carry the thread. We will.

Capture what you were doing before you leave. Open it on the next machine and continue.

Agent sync

Why: Skills and MCP configs shouldn't be per-machine.

Sync rules, skills, and MCP configs across Cursor, Claude, and other tools on your network.

Shared chat UI

Why: Optional shared history on the LAN.

Browser UI on your model server—chat history on the network, not trapped on one laptop.

Local supervisor

Why: Multi-agent workflows without sending data out.

YAML agents + RAG over your corpus. Your stack, your hardware, your data.

Model server → LAN → Cursor / Claude → LEMdesk sync → Next room

The journey

Three desks. One thread.

This is the story you tell when you demo it: leave home with context saved, arrive at the office with zero re-setup, update on the road.

Room 1 · Home

Deep work

Model server runs. One sync command rebuilds RAG and runbooks. You ship features with local docs in context.

Smart Handoff → saved
Room 2 · Office

Zero re-setup

Same repo, same LAN model URL, paste the handoff prompt. Cursor picks up mid-thread—your friends notice.

Agent context → loaded
Room 3 · Travel

Keep shipping

VPN or LAN extend. Menu bar shows desk health. Update handoff before you land.

Desk pack → updated

Onboarding

Live in four steps

No SaaS signup. No cloud account. Python + a model server on a trusted LAN—that's the whole requirement.

  1. Run models once

    One machine hosts your OpenAI-compatible API (Docker Model Runner on :12434 works great). Every other desk points at it.

  2. Sync the desk

    lemdesk_auto.py --fast rebuilds knowledge, RAG, runbooks, and a handoff template—your desk's memory on disk.

  3. Wire Cursor

    Set base URL to your server IP. Add LEMdesk MCP so agents search your local corpus, not the web.

  4. Hand off & move

    Before you switch rooms: Smart Handoff (Pro) or session handoff (Starter). Paste in the next Cursor. Same thread.

Positioning

Cloud desk vs LEMdesk

We're not competing with Cursor or Claude—we're the glue between your rooms and your local stack.

Typical cloud AI LEMdesk
Data stays on your LAN ✕ ✓
Same setup every room Partial ✓
Local doc search (MCP/RAG) ✕ ✓
Cross-room agent handoff ✕ ✓
Open source core ✕ ✓
Starter cost $$$ $0

Pro is $19/mo for menu bar, Smart Handoff, and dashboard—see pricing.

Who it's for

Built for a specific kind of developer

Multi-room builders

Home studio + office + laptop. You want one thread, not three cold starts.

Local-first engineers

Docker Model Runner, Ollama, or LAN endpoints—you refuse to rent your own GPU in the cloud.

Cursor / MCP power users

Skills, rules, doc search, agents—you need that config portable across machines.

Indie & side-project devs

Starter is free forever. Pro costs less than most SaaS seats—and you own the data.

Get started

Clone. Sync. Go.

Under five minutes on a trusted LAN with Python 3.10+ and a model server.

Terminal 
# Clone LEMdesk Starter (free sample)
git clone https://github.com/billrilea-lab/LemDesk.git
cd LemDesk
pip install -r requirements.txt

# Starter: one-shot sync
python3 lemdesk_auto.py --fast --skip-dmr-check

# Pro: menu bar + dashboard (after Starter sync)
python3 -m lemdesk_pro menubar

# Point Cursor → OpenAI Base URL
# http://<server-ip>:12434/engines/v1
Join Pro waitlist Clone Starter free See demo

FAQ

Common questions

Who is LEMdesk for?

Developers who work across multiple machines, run local or LAN models, and use Cursor or Claude with MCP. If you only use one laptop with cloud AI only, you probably don't need us. If you have three desks and hate re-setup, you do.

Is LEMdesk a cloud service?

No. Everything runs on your machines. LEMdev provides scripts and templates you run yourself.

Does it replace Cursor or Claude?

No. It wires your existing tools to your local model and your local knowledge.

Is it affiliated with Docker or any AI vendor?

No. LEMdesk is an independent project from LEMdev. You choose which third-party tools to use.

What's the difference between Starter and Pro?

Starter is the free sample kit—CLI scripts, MCP, manual handoff markdown. It proves local desk sync works. Pro is the daily driver: menu bar health score, Smart Handoff desk packs, local dashboard, and one-click sync. Same open core; Pro adds the polish you actually live in.

What does it cost?

Starter stays free forever. LEMdesk Pro is $19/mo (waitlist open)—founding members lock in that price.

Is chat history synced in Cursor?

Not natively. LEMdesk adds handoff files, optional shared browser chat, and config sync—not full Cursor cloud parity.

Stop re-setting up every room.

Try Starter tonight. Join Pro when you're ready to live in it.

Join Pro waitlist Watch the demo