Lorah
Download

Quickstart/Start here

Your first fifteen minutes.

Install, add one key, open a workspace, ask your team something real, approve one thing. Then you are in.

Fifteen minutesFree forever · fourteen days of Pro to startmacOS and Windows

Start here · the Quickstart

Install Lorah

Download from lorah.ai or the GitHub Releases page. No invite, no code, no account, no card, no waitlist. The macOS build is signed with an Apple Developer ID and notarized, and the Windows build is signed. Both name Yoram Golandsky as the publisher.

macOS. Drag it in.

Apple silicon or Intel, macOS 12 or later.

Lorah
Lorah
Applications
  1. Open Lorah-arm64.dmg, or Lorah-x64.dmg on an Intel machine.
  2. Drag the Lorah icon onto Applications, then open it from there.
  3. If macOS still blocks it, download again from lorah.ai or the Releases page and write to support@lorah.ai. Don't switch Gatekeeper off for it.

Windows. Run the setup.

Windows 10 or later, 64-bit.

User Account Control×

Do you want to allow this app to make changes to your device?

Lorah
Verified publisher: Yoram Golandsky
File origin: Downloaded from the Internet

YesNo
  1. Open Lorah.Setup.exe from your Downloads folder.
  2. Windows asks whether to allow it. Check that the publisher is Yoram Golandsky, then allow.
  3. A brand-new build can show a SmartScreen notice for a while. Same check: the publisher name. If it isn't ours, don't install, and write to support@lorah.ai.

Add your first key

Onboarding walks you through it the first time Lorah opens. Pick one provider, paste the key, done.

Connect a provider

One key is enough to start. You can add the others later, in Settings.

AnthropicOpenAIGeminiPerplexityxAIOllama, local
sk-ant-api03-••••••••••••••••••••••••••7Kq2Save key

Stored in your operating system's keychain. Sent only to the provider, never to Lorah.

Onboarding, connecting a provider. The key is stored in your operating system's keychain and sent only to that provider, never to Lorah's servers.

One key is enough.Anthropic, OpenAI, Gemini, Perplexity or xAI. Model usage is billed by the provider, at their price, on their invoice. Lorah adds nothing. An Anthropic or OpenAI key also runs Lorah's own jobs: proposing a team, Sync, and room decisions. Gemini, Perplexity and xAI keys run specialists only.
Or a local model.Ollama or LM Studio, through an OpenAI-compatible endpoint. No cloud charge for those calls, so Lorah labels them unpriced; a hosted compatible endpoint can still bill you. A local model runs Lorah's own jobs only after it passes a quality test in Settings; until then those jobs use your Anthropic or OpenAI key.
What you getFreePro
SpecialistsUnlimited, 3 active at a timeUnlimited, all active
Memory items50Unlimited
Workspaces2Unlimited
Documents in Materials10Unlimited
Sync, Remember this, new rooms, MCPPause when the trial endsIncluded

Fourteen days of Pro to start, no card. When it ends nothing is deleted: memory, materials, past rooms and history stay readable, and adding memory by hand still works. Upgrade any time in Settings → License & Billing; checkout and the billing portal open in your browser and land you back in Lorah. Model usage is billed by your provider on every plan.

Free and Pro. The fourteen days unlock everything; the limits are what stays after.

Open a starter workspace

Also part of onboarding. Pick a ready-made team with starter memory, or describe your work and let Lorah propose one.

Pick a team, or describe one.

5 specialists
Personal Board of Directors Strategy · innovation · ethics · operations · story
7 specialists
CISO AI Risk Operating Room Chief of staff · security architect · GRC · SOC · red team
4 specialists
The Product Team Product lead · engineer · designer · growth

Build your own

Tell Lorah what you need

Talk it through with Lorah. She asks, you answer, the team takes shape. A pasted brief works too.

Hi, I’m Lorah. In a sentence or two: what are you working on, and what should a team take off your plate? Or paste a project brief and I’ll work from that.
I’m building a…

Ready-made teams, or tell Lorah about your work and build one.

The workspace picker. Rename, add or remove anyone afterwards.

Run your first room

Open Rooms, pick two or more specialists, and ask something real, a question from this week, not a toy. Everyone answers side by side. This is where Lorah stops feeling like an app and starts feeling like a team.

“Debate the strongest and weakest parts of this plan.”
“What am I missing?”
“Research the facts, then have the critic challenge the conclusion.”

One question. Three ways to ask it.

How rooms work
You

Timer only at launch, or tasting notes too?

asked three ways ↓

Ask one specialist

MayaProduct Lead
Timer only. One habit at launch.
And if roasters ask for notes on day one?
MayaProduct Lead
Give them a date. People wait for a date, not for a maybe.
Draft that line.
MayaProduct LeadIn memory
Roast timer now. Tasting notes in the next update.

Switch to Jack. The thread comes with you.

Ask the room

MayaProduct LeadDONE
Timer only. One habit at launch.
JackBusiness StrategistDONE
Both. Notes are what brings people back.
SophieBrand & ContentDONE
Whichever we pick, say what’s missing.
SynthesisThe roomSave to memory
Disagreement. Maya says timer only. Jack says timer and notes.

Nobody saw the others’ answers. You decide.

Work together

MayaProduct Lead
Holds, but the page needs a date for notes.
Jack, make the case for notes.
JackBusiness Strategist
Notes are the return visit. Ship a light version.
Sophie, write it.
SophieBrand & ContentAdd to Tasks
Roast timer now. Tasting notes in the next update.

They take turns. You steer between them.

Ask one specialist, ask the room apart, or put them to work together.

One question, three ways to ask it: one specialist, the whole room apart, or the team working together. Prefer a quiet start? Open any specialist one to one and ask them to draft something in their area, or what they think is missing.

Approve one thing, then come back tomorrow

Something useful comes out of your first room. Save to memory under the conclusion keeps it as a decision. Back in a chat, press Wrap up session, check the summary and press Save to memory: the facts and decisions land in Review. Approve one. Close Lorah. Open it tomorrow. The team remembers what you approved, and the next conversation starts from context, not from scratch. The Memory guide covers Review and what happens after.

Review: keep one, edit one, reject one. Nothing enters memory without your yes, and the first extraction can take thirty to sixty seconds.

What it costs you

Every reply carries its price, and the workspace has a ceiling.

Settings · Budget

Monthly workspace budget

$50Your ceiling, not a payment. One number, one workspace, one month.
This month $18.70Warns at 80% · pauses at 100%
Today$1.42
Last reply$0.03 · claude-opus-4-6
Local modelsunpriced, outside the budget
Budget at 82%. $41.10 of $50 this month.
Budget reached

New sends are paused for this workspace. You can still send this one.

Send anywayRaise the capNot now
MayaProduct Lead
Give them a date. People wait for a date, not for a maybe.
$0.03 · checkedapproximateunpricedhow sure the price is
Receipt. 1.2k of 200k tokens · memory: 12 items fit · web off.

Settings, Budget: the cap, month to date, the warning at 80% as a banner across the app, and the pause at 100% with Send anyway. Spend is tracked by model, by specialist and by surface. Before a room or a decision runs, the door shows an estimate and can swap in a cheaper lineup; larger runs get an estimate the same way. Every reply carries a cost chip that says how sure it is: checked against the provider's price list, approximate, or unpriced for a local model.

Then the guides

Five guides, each with a short walkthrough video.

01ChatOne specialist at a time, and every control on the box you type in.02RoomsAsk everyone at once, or have them work it through one at a time.03MemoryThe shared brain: what it holds, how things get in, how to correct them.04MaterialsDocuments your team can cite, and what "assigned" really means.05MCP setupLet Claude Desktop, Cursor or Claude Code read your memory. A reference.
Local-first.Your project is stored on your computer. Model requests go to the providers or endpoints you configured. Lorah also uses the network for licensing and updates, and for web access or connected tools you use. Crash reporting is optional and off by default.
Cost is visible.Every priced reply carries a price. The budget above warns and pauses new priced work at the threshold you set; you can override it, and estimates can differ from the final charge. Local work is unpriced and sits outside it.
Pro when you want it.Free stays free. Pro is $19 a month or $190 a year, and cancelling never touches your workspace.
NextChatOne specialist at a time, and the box you type in.RelatedPricingWhat the fourteen days unlock, and what pauses after.Stuck?Write to ussupport@lorah.ai, or message Yoram on WhatsApp. Lorah is built by a small team and we read everything. Discord and a bug tracker are coming.