Capture intent
A user writes a search in plain language: play tennis, find a running group, discuss a project, meet a supplier, join a table.
Semantic intent matchmaking
Say what you want to talk about. LMGram finds the right people.
LMGram connects people through live intentions: a topic to discuss, a plan to coordinate, a partner to find, an idea to test, a group to join, or a question that needs another human.
Semantic matching
Intent-based chat
Nearby signals
Intent rooms
ChatGPT + CLI
Live intent
What are you looking for?
I want to find someone nearby to talk about building an AI social product
5 matches
meaning + context
AI Social Product Circle
People with compatible intentions can start direct chats or coordinate as a group.
Maya I am testing onboarding for an AI-native community app.
Alex I want feedback on semantic matching and cold start.
Priya I can compare this with how creator networks grow.
Open to compatible intents
What LMGram is
A matchmaking engine for intentions, not a directory of profiles.
Most social products start by asking who you are. LMGram starts with what you want now. The system turns natural language into intent signals, compares them against active intentions, and creates the right conversation when two or more people should talk.
Understand meaning
LMGram uses semantic embeddings, intent cards, lexical overlap, complementarity, sensitivity checks and optional location.
Find compatible people
The match is not just keyword similarity. It looks for overlapping, compatible or complementary intentions.
Open chat
When there is a useful match, people can start lightweight chat with text, photos, video, audio and notifications.
Create rooms
When the opportunity is group-shaped, compatible people can join an intent room instead of fragmented one-to-one chats.
How it works
From a sentence to a human conversation.
- The user says what they want"I need someone to practice tennis with after work", "I want to debate this startup idea", "I am looking for a running group near me".
- LMGram models the intentThe platform creates semantic and structured signals while keeping the original phrase available for context.
- Active intentions are comparedThe engine scores similarity, compatibility, complementarity, location and conflict risk.
- The right conversation opensA direct chat or an intent room appears when people have a real reason to talk.
The use case is the intent
LMGram can organize almost any reason people should talk.
The product is not limited to business events. A fair, conference or summit is a powerful environment because many people arrive at once, but the core primitive is broader: matching human intentions.
Sports and activities
Find someone for tennis, running, hiking, cycling, gym sessions, pickup games or a spontaneous weekend plan.
Ideas and projects
Find builders, reviewers, early users, mentors or professionals willing to discuss a concrete project or decision.
Local conversations
Meet people nearby who want to talk about the same topic, visit the same place or coordinate the same activity.
Learning and practice
Match people who want to practice languages, study a topic, exchange feedback or teach and learn complementary skills.
Professional discovery
Connect founders, operators, candidates, investors, customers, suppliers or consultants around explicit needs.
Group formation
Turn scattered compatible intentions into a group room for dinners, workshops, roundtables, teams or shared plans.
Detailed use case: events
Events are where LMGram can feel immediately obvious.
At an expo, fair, summit or conference, everyone is already in the same domain and the window of opportunity is short. LMGram helps attendees stop browsing static lists and start expressing what they actually need before the event ends.
Expo Intent Hub
buyers, suppliers, partners, side meetings
BBuyerlooking for sustainable packaging under a specific MOQ
SSupplieroffering compostable packaging for small beverage brands
PPartnerwants to co-host a retail distribution roundtable
Better than badge scanning
People do not just collect contacts. They publish intent: what they need, what they offer, who they should talk to and why.
Works with density and time pressure
Events compress many compatible people into the same place for a short period. LMGram makes that hidden demand visible.
Supports groups, not only pairs
A single intent can produce a buyer-supplier chat, a founder roundtable, a dinner group or a thematic room.
ChatGPT, API and lmgram-cli
LMGram can be triggered by people or by the tools working with them.
A user can create an intent from the mobile app, from ChatGPT, from an API-backed integration or from `lmgram-cli`. That makes LMGram a human routing layer for AI agents and workflows that eventually need another person.
$ lmgram login
Approve this agent in LMGram mobile
$ lmgram intent create "Need a product operator to review onboarding"
Intent created: intent_8x41
Expires: 72h
$ lmgram matches list --intent intent_8x41
3 compatible people found
$ lmgram match accept match_24b
Chat openedMobile-first
Users can express intents, receive matches, chat, send media and use optional location signals from the mobile app.
ChatGPT connector
ChatGPT can ask for consent, summarize a need and create a minimized LMGram match request through OAuth.
Agent and API workflows
Agents can detect when a workflow needs a real human and create an explicit intent tied to the user's LMGram account.
User intent
Semantic engine
Compatible people
Chat or room
Coordination
Trust and control
Intent-based matching still needs clear safety controls.
LMGram is designed around opt-in matching, direct control and progressive conversation. Users can choose when to search, whether location participates, when to chat and when to leave or report a conversation.
Opt-in intentUsers create searches intentionally. LMGram does not require public bios to discover useful people.
Optional locationLocation can boost nearby matches when useful, but the user can disable it.
Blocking and reportingUsers can block, report and remove themselves from unsafe or irrelevant conversations.
Admin tuningScoring and gating can be adjusted without redeploying when real-world tests reveal better behavior.
One engine, many surfaces