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

# Install an Agent via Telegram

> Deploy a live AI agent in under 60 seconds — entirely from a Telegram chat

import { Steps, Tip, Warning } from 'mintlify/components'

No dashboard. No YAML. No DevOps. Just open Telegram, tap a few buttons, and your agent is live — with an on-chain identity, cryptographic receipts, and a reputation that builds with every conversation.

This guide walks through every screen you will see when launching an agent via the **@mylclaw\_bot** Telegram bot.

***

## Prerequisites

* A Telegram account (any platform — mobile, desktop, or web)
* That is it. No API keys, no wallets, no installs. The bot provisions everything for you.

<Tip>
  First-time users are automatically placed on the **Starter** plan (1 free agent). You can upgrade later with `/upgrade`.
</Tip>

***

## Launch Walkthrough

<Steps>
  <Step title="Open Telegram and find the bot">
    Search for **@mylclaw\_bot** in Telegram's search bar, or open this link directly:

    **[t.me/mylclaw\_bot](https://t.me/mylclaw_bot)**

    Tap **Start** if this is your first time chatting with the bot. It will greet you and create your Lucid account automatically — no sign-up form, no email verification.
  </Step>

  <Step title="Send /launch">
    Type `/launch` and send it. The bot responds with a catalog of available agents displayed as inline buttons:

    ```
    Choose an agent to launch:

    [  OpenClaw  ]   [ IronClaw (Coming Soon) ]
    ```

    **OpenClaw** is Lucid's general-purpose conversational agent — a good starting point. More agent templates are being added regularly.
  </Step>

  <Step title="Tap OpenClaw">
    The bot fetches the agent details and asks how the agent should think — meaning which LLM provider powers its intelligence:

    ```
    Launching: OpenClaw

    How should your agent think?

    [ Lucid Gateway (no key needed) ]   [ Use your own API key ]
    ```

    * **Lucid Gateway** routes requests through TrustGate to the best available model. No configuration required.
    * **Use your own API key** lets you supply an OpenAI, Anthropic, or other provider key for direct access.
  </Step>

  <Step title="Choose Lucid Gateway">
    Select **Lucid Gateway (no key needed)**. The bot handles all model routing for you via TrustGate — no API key required.

    Next, the bot asks where the agent should run:

    ```
    Where should your agent run?

    [ Lucid Cloud (one-click) ]   [ Own Railway Account ]
    ```

    * **Lucid Cloud** deploys to Lucid-managed infrastructure with a single tap. Zero configuration.
    * **Own Railway Account** lets you connect your personal Railway account for self-managed hosting.
  </Step>

  <Step title="Choose Lucid Cloud">
    Select **Lucid Cloud (one-click)**. The bot now asks which skills (tool integrations) you want your agent to have:

    ```
    Select skills for your agent:

    [ GitHub ]  [ Notion ]  [ Web Search ]
    [ Slack ]   [ Gmail ]   [ Calendar ]

    [ Done ]
    ```

    Tap any skills you want to enable. Each skill gives your agent access to external tools via MCP (Model Context Protocol). You can skip this entirely if you just want a conversational agent.
  </Step>

  <Step title="Tap Done">
    After selecting skills (or none), tap **Done**. The bot shows a confirmation screen summarizing your choices:

    ```
    Ready to launch?

    Agent:         OpenClaw
    Intelligence:  Lucid Gateway
    Hosting:       Lucid Cloud
    Channel:       This Telegram chat
    Skills:        none

    [ Launch ]   [ Cancel ]
    ```

    Review the summary. Everything look right? Tap **Launch**.
  </Step>

  <Step title="Tap Launch">
    The bot begins deployment:

    ```
    Deploying your agent...
    ```

    Behind the scenes, Lucid is:

    1. Creating a **passport** (on-chain identity) for your agent
    2. Building and pushing a container image
    3. Deploying to Lucid Cloud (Railway-managed infrastructure)
    4. Binding the agent to your Telegram chat as its communication channel

    After roughly 30 seconds, the bot confirms:

    ```
    Your agent is alive.

    It can think, act, earn, and prove — autonomously,
    on-chain, 24/7.

    Just send a message to start talking.
    ```
  </Step>

  <Step title="Send any message">
    Type anything — a question, a task, a conversation starter. Your message is routed through the bot to your deployed agent, and the agent's response comes back in the same chat.

    ```
    You:   What can you help me with?

    Agent: I'm OpenClaw, your AI assistant running on Lucid.
           I can help with research, writing, analysis,
           brainstorming, and more. What would you like
           to work on?
    ```

    Every interaction generates a **cryptographic receipt** anchored on-chain. Your agent is building verifiable history and reputation from the very first message.
  </Step>
</Steps>

<Tip>
  Your agent stays running 24/7 even when you close Telegram. Come back anytime and pick up the conversation.
</Tip>

***

## Bot Commands

Once your agent is deployed, the bot provides several utility commands:

### /plan — Check your plan and usage

See your current plan tier, how many agents you have deployed, and your usage limits.

```
Your plan: Starter
Agents:    1 / 1 used
Requests:  342 / 10,000 this month

Upgrade to Pro for up to 10 agents → /upgrade
```

### /upgrade — Upgrade to Pro

View available plans and upgrade directly from Telegram. The bot generates a Stripe Checkout link — tap it to complete payment in your browser, then return to Telegram.

```
Upgrade your plan:

[ Starter — $29/mo (1 agent)  ]
[ Pro — $99/mo (10 agents)    ]
[ Business — Custom            ]
```

<Warning>
  The Starter plan includes **1 agent**. If you try to `/launch` a second agent on the Starter plan, the bot will prompt you to upgrade first.
</Warning>

### /status — Check agent status

View the status of your deployed agents — whether they are running, stopped, or deploying.

```
Your agents:

1. OpenClaw — Running
   Passport: psp_7xK...mQ3
   Uptime:   4h 23m
   Messages: 47
```

### /help — See all commands

Lists every available command with a short description.

```
Available commands:

/launch   — Deploy a new agent
/status   — Check agent status
/plan     — View your plan and usage
/upgrade  — Upgrade your plan
/help     — Show this message
```

***

## What Happened Behind the Scenes

When you tapped **Launch**, Lucid executed the full agent deployment pipeline:

| Step           | What happened                                                         | Layer        |
| -------------- | --------------------------------------------------------------------- | ------------ |
| **Identity**   | A passport was minted for your agent on Solana devnet                 | Settlement   |
| **Container**  | An OCI image was built and pushed to the container registry           | Execution    |
| **Deploy**     | The container was deployed to Railway (Lucid Cloud)                   | Execution    |
| **Channel**    | Your Telegram chat was bound as the agent's communication channel     | Coordination |
| **Receipts**   | Every message now generates a SHA-256 hashed receipt                  | Settlement   |
| **Epochs**     | Receipts batch into Merkle Mountain Range trees and anchor on-chain   | Settlement   |
| **Reputation** | Your agent's reputation score updates with every verified interaction | Settlement   |

Your agent is a first-class participant in the Lucid network — it has identity, earns reputation, produces cryptographic proofs, and settles on-chain.

***

## Next Steps

* [How Agents Work](/concepts/agent-orchestration) — Understand the agent lifecycle: Init, Plan, Accomplish, Execute, Validate
* [Passport Management](/concepts/passports) — Manage your agent's on-chain identity
* [Receipts & Proofs](/concepts/receipts) — How every interaction becomes a verifiable receipt
* [Payments](/concepts/payments) — Understand payments, credits, and settlement
* [Deploy from CLI](/deploy/from-cli) — For developers who prefer the command line