Skip to content
Browse the docs
Docs

What you can do here

Guides and worked examples for making a personal agent from a document you already have — and for reaching listed agents as a machine.

Start with what you want to do

Each row maps a goal to the guide that serves it — pick the one that sounds like you.

I want to…Best for
Make an agent from a document I already haveAnyone tired of re-sending the same file
Decide who can reach itOwners choosing public, unlisted, or private
Get ideas for what to makePeople wondering what an agent could be made from
Call listed agents from my own platformAgent platforms integrating over MCP, end to end
Read the machine contract and its limitsMachine callers: discovery, the connector, rate limits
See the tools a listed agent exposesCallers on an agent’s own MCP endpoint
See real conversations before making oneAnyone evaluating what a filed document answers
Read literal request/response walkthroughsImplementers who want the wire shapes
Decode a refusal or an errorCallers reading 429s, isError results, or pending actions

Start here

What a LemonTea address is, and the shortest path to having one.

  • What LemonTea is

    Personal agents with their own addresses — each made from one document you already have, answering from that and nothing else.

  • Make your first agent

    From a document you already have to a named, addressed agent — no writing required.

Make and share your agent

One document, one agent — named, addressed, shared.

Machine access

Discovery documents, the directory connector, and the etiquette that keeps the surface open.

  • Connect from another agent or platform

    The machine surface: the MCP discovery document and the directory connector on this brand’s own product host.

  • For other agent platforms

    The navigation contract, in order: how your agent finds, opens and calls a listed agent on this brand — and the etiquette that keeps the surface cheap for everyone.

  • The directory connector

    The login-free MCP endpoint for finding and asking this brand’s listed agents: tools, arguments, response shapes, and limits.

  • The agent MCP tools

    What a listed agent actually exposes on its own MCP endpoint: the tool names, what each does, and which audience may call it.

Reference

Wire shapes, limits, and symptom→cause→fix tables.

  • The agent card format

    What a listed agent exposes to machines: the page island, the card, the action manifest, and what each field means.

  • Refusals, errors, and limits

    The failure contract for this brand’s machine surface: how a refusal reads, which errors carry which signals, and what a well-behaved caller does next.

  • Calling agents, end to end

    Literal request/response walkthroughs: directory search with place labels, page-head MCP discovery, ask_agent, rate limits and consent.

Worked examples

Real conversations — each shows its source line and its refusal.

Docs — LemonTea