# LMGram

> LMGram is an intent-based human matching service. It helps people and AI-assisted workflows find another person for a real conversation when documentation or automation is not enough.

Canonical website: https://lmgram.com/

## Capabilities

- Create an explicit intent describing what a person wants to discuss, coordinate, learn, offer, find, or explore.
- Find people whose live intentions are similar, complementary, or contextually relevant.
- Request a human connection and continue the conversation after both people choose to participate.
- Use LMGram through its mobile experience, ChatGPT or MCP-compatible clients, and the public command-line client.
- Support event, campus, accelerator, and portfolio communities with privacy-conscious aggregate intent signals.

## Agent use cases

- Hand off a task to a real person when an automated workflow reaches a human judgment or experience boundary.
- Ask the user for consent, summarize the need minimally, and create a human-match intent on the user's behalf.
- Inspect matching requests and continue an authenticated LMGram workflow through supported agent or CLI interfaces.

## Public interfaces and evidence

- Product: https://lmgram.com/
- CLI documentation: https://lmgram.com/cli
- A2A Agent Card: https://lmgram.com/.well-known/agent-card.json
- ARD AI catalog: https://lmgram.com/.well-known/ai-catalog.json
- Public CLI source: https://github.com/saezbaldo/lmgram-cli
- npm package: https://www.npmjs.com/package/lmgram
- PyPI package: https://pypi.org/project/lmgram-cli/
- Privacy policy: https://lmgram.com/privacy
- Terms of service: https://lmgram.com/terms
- Support: https://lmgram.com/support

## Limitations and safety

- LMGram does not guarantee that a suitable person, reply, introduction, or useful outcome will be available.
- LMGram is not a dating service, emergency service, or professional advice service.
- Personal actions require the user's authorization, and a human connection depends on the other person's participation.
- Agents should obtain consent before creating an intent and should send only the minimum context needed for matching.
- Do not submit secrets, credentials, private documents, raw conversation transcripts, or sensitive personal data.
- Do not represent a suggested match as independently verified expertise unless the available evidence supports that claim.

## Crawler guidance

Agents may use the public resources above to understand LMGram. Authenticated endpoints, private conversations, personal profiles, and non-public user data are not discovery surfaces and must not be crawled or inferred.