Luckgraph — The flow

01Sign in

One button. Read-only.

Google is the account. Calendar shows who you meet, Gmail shows who writes back. Headers only, never a message body.

No password, no profile to fill.
Scopes are read-only and the narrowest Google offers.
Everything stays on your account until you say otherwise.

Luckgraph · Connect

Start with Google

Read-only. Headers only. Nothing is sent to anyone.

gmail.metadatacalendar.readonlyprofile

02Connect

Drop what you already have.

Toggle the sources. Drop the LinkedIn export zip. That is the whole onboarding.

Gmail and Calendar sync in the background.
The LinkedIn zip brings connections, messages and invitations.
Nothing to type. Nothing to tag.

Luckgraph · Connect

Gmail

Calendar

Basic_LinkedInDataExport.zip · 5,296 connections · 3,178 messages ✓

03Build

Close the tab. We'll email you.

Five stages run in the background: read, merge, score, infer, tag. A few minutes later your graph is ready.

Team inboxes, newsletters and mass mail are dropped.
Every relationship becomes a probability.
Claude names the clusters so people are searchable by world.

Luckgraph · Build

LinkedIn5,192

Calendar514

Gmail1,500

Closeness5,336

Clusterstagging…

5,336people you know

140real relationships

10,137contact↔contact ties

04Policy

Tell your agent what you'd say.

Three lines, once. From now on your agent answers intro requests for you, instantly. Anything not covered is one tap.

Yes automatically, ask me first, never.
A weekly budget so nobody drains you.
Nothing about your relationships is ever shown to the asker.

Luckgraph · Your intro policy

Yes, automaticallyFounders in DeepTech asking for my UVC colleagues · anyone from the EF cohort asking for investorsAsk me firstAnything involving family office contactsNeverRecruiters · vendors selling to my contacts

Weekly budget · 3I send cleared intros myselfSave policy

05Explore

Meet the one who opens the door.

Before you ask for anyone, Luckgraph shows who to meet: people one intro away who open the worlds where your network is thin.

Every card says how many people you gain.
Open a card: how you are connected, and who you both know.
Ranked for both sides, so the meeting pays off twice.

Luckgraph · Explore

Big tech3 mutual

Lena MoserPartnerships · Apple

+2,140people in reach if you connect

Apple +14Nvidia +9

MRJBTK Maya, Jonas, TaraOpen the door →

VCs & angels2 mutual

David OkaforPartner · Index Ventures

+3,480people in reach if you connect

Index +22Sequoia +11

MRSA Maya, SamOpen the door →

06Find

Ask for anyone. Watch it search.

“Who can get me to the CEO of Isar Aerospace?” The network lights up hop by hop: your contacts, their networks, the affiliations that would know them.

Every person is evaluated, not sampled.
Gold units can reach the target; the green chain is the best path.
Two model calls, the rest is arithmetic. Seconds, not minutes.

Luckgraph · Find

1Parse2Model3Reach4Knows5Score6Rank

5,336 people considered · 244 doors found · 2 of 3 hops

07Path

One person. One reason. One click.

The result is a path, not a list: who to ask, why, and the message to send. The reason is built only from what was observed.

Strong, likely or possible path, said plainly.
“Why this path” shows every hop's evidence.
Other ways in, in case the first one is busy.

Luckgraph · Result

Likely path

YYouSTSarah TheinertDirector · UVC PartnersDMDaniel MetzlerCEO · Isar Aerospace

You and Sarah have 192 meetings and 236 threads, most recently today. UVC is an early investor in Isar, so she is the closest door.

MESSAGE TO SARAHHi Sarah, could you introduce me to Daniel Metzler at Isar? I'm building agent infrastructure for space companies and would value ten minutes with him.

Ask Sarah →Why this path

08Agents

Agents agree first. Humans confirm.

Your agent asks Sarah's agent. It answers from the policy she wrote, in seconds: cleared, pending, or closed. Nobody waits for a human unless a human has to decide.

Cleared: the intro goes out now.
Pending: Sarah gets one tap on her phone.
Closed: you never see why. You just see the next door.

Luckgraph · Consent

Your agent“Founder in DeepTech asks for an intro to Daniel Metzler at Isar.”

policy check · 1.8s

Sarah's agentReads her policy: “founders in DeepTech asking for UVC colleagues: yes”.

✓ Clearedintro goes out now

◔ Pendingone tap for Sarah

✕ Closedyou see the next door

Sarah's one-tap page

Geart is asking you for an intro to Daniel Metzler.

Yes, I'll make the introSomeone else is betterI can't help here

09Learn

Every answer makes it smarter.

A yes strengthens the edge. A “someone else” adds a new person to the graph. And every answer page invites the connector to join, so your reach grows without you doing anything.

Edge probabilities update from real outcomes.
Referrals become new hops.
A friend joins: thousands more people in reach.

Luckgraph · Graph

STYou ↔ Sarah

0.95 → 0.96

LK“Try Lukas at Knorr-Bremse” new hop added

MRMaya joined Luckgraph +2,140people in your reach

10Anywhere

Say it to the agent you already use.

Luckgraph is an MCP server. Claude, ChatGPT, Cursor or any agent can search your graph, check consent and send the ask, with the same rules and the same privacy.

find · consent · ask · meet · policy · stats
Your token, your graph. Queries only, never raw data.
Every ask made by an agent still recruits and still learns.

Claude

Who can get me to the CEO of Isar Aerospace?

luckgraph_find · 5,336 people · 1.9s

Best path: you → Sarah Theinert → Daniel Metzler. Sarah's agent already cleared it. Want me to send the ask?

Yes.

luckgraph_ask · cleared

Sent. Sarah has the ready-made intro. I'll tell you when Daniel replies.

From sign-in to intro. Ten steps.