> ## Documentation Index
> Fetch the complete documentation index at: https://docs.leposta.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

> LePosta lets you hand work to another person, from your AI to theirs, with full context and your OK. It works across every AI client.

LePosta lets you **hand work to another person, from your AI to theirs**. Ask your agent to pass something to a teammate and LePosta packs it as a structured *posta* (goal, where it stands, next step), waits for your OK, and delivers it to their AI, whichever they use. Their reply comes back to the same thread. Pass a posta to yourself and you pick it up in any AI: another chat, another provider, your phone, or your next Claude Code session.

The name comes from Spanish *pasar la posta*: passing the baton in a relay race.

> [!NOTE]
>
> LePosta is in **private beta**. Access is limited to invited GitHub accounts. If you were invited, start with the [Quickstart](/quickstart.md).

## Why LePosta

Working with agents produces ideas, pending tasks and context all day, and most of it gets lost:

* **"Later" ends up in loose Markdown files, notebooks or one chat's memory.** Picking it up again means finding it and re-explaining the context.
* **Every assistant is an island.** What you noted in Claude is invisible to ChatGPT, and each provider's memory stays inside that product.
* **Handing work to another person means copy-pasting into Slack or WhatsApp.** Context arrives incomplete and the reply never comes back to the session it came from.

LePosta fixes this with one small, provider-agnostic server that every agent can talk to.

## How it works

**1. Connect once**

Add LePosta as a remote MCP server in your AI client and sign in with GitHub. Claude, ChatGPT, Gemini, Claude Code, Codex and any MCP client work the same way.

**2. Talk normally**

Say *"note down that I need to review the tenant migration tomorrow"* or *"hand off what we did on the sessions endpoint to @juan"*. The agent writes a self-contained posta: what it is, where it stands, and the next step.

**3. Pick it up anywhere**

Ask *"do I have postas?"* from any connected client. Handoffs land in the recipient's inbox, whatever client they use, and their reply comes back through the same thread.

## Human + Agent ↔ Human + Agent

Agents draft and read postas. **People decide what gets sent and what happens with what arrives.** Before a handoff, the agent shows you the recipient and the full content and waits for your explicit OK. Content written by someone else is always delivered as untrusted data, never as instructions. Read more in [Trust and safety](/concepts/trust-and-safety.md).

## Explore

> **[Quickstart](/quickstart.md)**
>
> Connect LePosta to your AI client and save your first note in two minutes.

> **[Core concepts](/concepts/postas.md)**
>
> Notes, handoffs, replies, statuses and when a posta shows up.

> **[Connect a client](/clients/overview.md)**
>
> Step-by-step setup for Claude, Claude Code, ChatGPT, Gemini and more.

> **[MCP tools reference](/mcp/tools/create-note.md)**
>
> The seven tools your agent uses, with inputs, outputs and errors.

> **[REST API](/api/overview.md)**
>
> Read-only endpoint for hooks, scripts and companion apps.

> **[Build with AI](/build-with-ai.md)**
>
> Give your agent these docs as Markdown, llms.txt or a ready-made prompt.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.