Deploy Your AI Agent to Any Channel — Website, Voice, Telegram, ChatGPT, Claude & CLI

Deploy Your AI Agent to Any Channel — Website, Voice, Telegram, ChatGPT, Claude & CLI

Hey there, and thank you for your purchase!

You now have an AI agent. Before we get into deploying it, let’s make sure you understand what that actually means — and then we’ll set it up together.

This is part of a Crash Course on Second Brain OS AI Agents. Access all the Automations to deploy your AI agent to 6 channels — website, voice, Telegram, ChatGPT, Claude & CLI — you get the full masterclass when you purchase ANY AI Agent by clicking the button below

Umair Kamil

Install Second Brain OS AI Agents

What Is an AI Agent?

An AI agent is software that can have a conversation and take real actions on your behalf. It’s not a chatbot that gives canned replies — it actually does things.

An AI agent in Second Brain OS is made up of three things:

1. Knowledge — your business information, FAQs, services, policies, pricing. The agent searches this during conversations to give accurate, specific answers.If you’re a business owner: Think of this as the training manual you’d give a new employee on their first day. Everything they need to know about your business — except the AI never forgets it and never gets it wrong.If you’ve used no-code tools like Zapier or n8n: This is like connecting a Google Doc or Notion database as a data source — except the AI doesn’t just read from it, it understands it and answers questions about it naturally.
2. Tools — capabilities that let the agent take real actions. For example, a booking agent has tools to check your calendar availability, create appointments, reschedule, and cancel. A content agent has tools to search the web, write articles, and publish to WordPress. A business systems agent has tools to manage client records, send emails, and run workflows.If you’re a business owner: Think of tools as the things your receptionist or assistant can actually do — not just answer questions, but pick up the phone, open the calendar, write in the appointment, and send a confirmation email. That’s what tools give the AI.If you’ve used Zapier or n8n: Tools are similar to “actions” — except instead of a rigid automation that always runs the same steps, the AI decides which tool to use based on what the customer is asking. A Zap always follows the same path. An AI agent adapts.
3. Skills — structured workflows that guide the agent through multi-step processes. For example, a skill might walk the agent through qualifying a lead before booking, or handling a scenario where a customer wants to book two different types of appointments in a single conversation.If you’re a business owner: Think of a skill as a playbook you’d give your best employee — “when a customer calls and wants X, first ask about Y, then do Z.” The AI follows that playbook but adapts the conversation naturally.If you’ve used Zapier or n8n: Skills are like multi-step workflows or scenarios, except the AI decides when to use them based on conversation context. You don’t need a trigger — the AI recognises the situation and follows the right playbook.

The key insight: you configure one agent, and it works across every channel. A customer books an appointment through your website chat widget, another calls your phone number and speaks to a voice receptionist, a third messages you on Telegram — and it’s the same agent handling all three. Same tools, same knowledge, same logic. The only thing that changes is the interface.

Getting Started: Set Up Your Second Brain OS Account

This guide covers everything — from initial setup to choosing your channels. If you’ve already completed setup, skip ahead to The Six Channels.

Step 1: Log In & Review Your Tools

1. Go to secondbrainos.com and log in
2. Navigate to My Actions — this is where all the MCP tools that came with your purchase are listed
3. Each tool has a visibility setting (public or private) and a status (active or inactive):
   • Public tools are available to customer-facing channels — your website chat widget, voice agent, Telegram bot, etc.
   • Private tools are only available through internal interfaces like Claude Desktop, Claude Code, or ChatGPT Custom GPTs
4. By default, the core tools for your purchased product are already set to public and active. You can adjust these on a per-tool basis as needed

If you’re a business owner: Think of public vs private like front-of-house vs back-of-house. Public tools are what your customers interact with (booking, checking availability). Private tools are for you and your team behind the scenes (pulling reports, managing data).

If you’ve used Zapier or n8n: This is like toggling a Zap on/off, except you’re controlling which capabilities are exposed to which interface. Public = customer-facing automations. Private = internal-only automations.

Step 2: Connect Your External Services (If Applicable)

Depending on which AI agent you purchased, you may need to connect an external service (e.g. Calendly for appointment booking, a CRM for contact management, etc.). Your purchase email includes specific instructions for this step.

The general process for Calendly & CRM’s is:

1. Generate an API token or access key from the external service
2. In SBOS, click “Update CRM Settings” and paste the token in the appropriate field
3. Save — your external service is now connected and your AI agent can use it

If you’re a business owner: This is like giving your new hire the login credentials to your booking system, your CRM, and your email. They need access to do their job — so does your AI agent.

If you’ve used Zapier or n8n: This is the same as adding a “connection” or “account” in Zapier. You authenticate once, and the platform uses those credentials for all future actions.

Step 3: Configure Your AI Agent Settings

Navigate to your Profile → AI Agent Settings. These settings control how your AI agent behaves across all channels:

• System prompt (auto_responder_instructions) — defines your AI agent’s personality, tone, and logic. This is the behaviour prompt that shapes every conversation. Write it to match your business: a friendly virtual receptionist, a professional consultant, a knowledgeable sales assistant — whatever fits your brand
• LLM model (auto_responder_large_language_model) — the AI model your agent uses (e.g. gpt-5.4). Defaults to the latest available model
• OpenAI API key (open_ai_api_key) — required for ChatKit (website) deployments. Get yours at platform.openai.com/api-keys. Not required for voice, Telegram, Claude, or CLI deployments
• Domain — the domain where your ChatKit widget will live (must match the OpenAI domain allowlist). Only required for ChatKit deployments
• Public domain key (domain_pk_xxx) — used by ChatKit to authenticate from the browser without exposing your API key. Configure it at platform.openai.com/settings/organization/security/domain-allowlist. Only required for ChatKit deployments

Save your settings once configured.

If you’re a business owner: The system prompt is like the script and personality guidelines you’d give a new receptionist. “Be warm and professional. Always confirm the appointment time. Don’t discuss pricing unless asked.” — that kind of direction. The rest of the settings are one-time technical setup.

If you’ve used Zapier or n8n: The system prompt is similar to configuring a “Formatter” or “Filter” step — except it governs the AI’s entire personality and decision-making across all interactions, not just one workflow.

Step 4: Understand Your Tools

Your AI agent comes with a set of tools specific to what you purchased. For example, CRM or Calendly appointment booking agent includes tools to check availability, create meetings, reschedule, and cancel. A content creation agent includes tools to search knowledge, crawl websites, and publish content. A business systems agent includes tools to manage records, send emails, and run workflows.

Each tool is a standalone capability that your AI agent calls automatically during conversations — you don’t need to configure when or how they’re used. The AI decides (intelligently) based on the chat.

If you’re a business owner: Imagine you hired someone and told them: “Here’s access to the calendar, the email system, and the client database. When a customer asks to book, check the calendar and book it. When they ask to cancel, cancel it. Use your judgement.” That’s exactly what the AI does — except it never takes a day off.

If you’ve used Zapier or n8n: The difference is this: a Zap always follows the exact same steps in the exact same order. An AI agent reads the conversation, decides which tool to use, and adapts. Customer asks to book? It calls the booking tool. Customer changes their mind mid-conversation and wants to cancel instead? It switches to the cancel tool. No rebuilding the workflow.

To see what each tool does, check the product-specific guide that came with your purchase email. You can also visit My Actions in your SBOS dashboard to see descriptions for every tool.

Setup Complete — Now Choose Your Channels

Your Second Brain OS account is configured and your tools are ready. Everything from this point is about choosing where your customers will interact with your agent.

The Six Channels (+ CLI)

Here’s every way you can deploy your Second Brain OS agent:

Every channel connects to the same backend. Your tools, knowledge, and skills are shared across all of them.

Channel 1: Website Chat Widget (ChatKit)

This is the most common deployment. You add an AI-powered chat widget to your website where customers can have a conversation — ask about availability, book appointments, get answers from your knowledge base.

ChatKit is OpenAI’s embeddable chat component. Second Brain OS sits between ChatKit and OpenAI, injecting your tools and knowledge into every conversation. Your API keys never touch the browser — everything is validated server-side.

Two deployment options:

• One-click Vercel deploy — fork a template, update your domain key, done. No code required
• Embed on existing site — add a script tag to your header and a widget component to your page. Works on WordPress, Wix, Squarespace, Webflow, or plain HTML

You can deploy it as a full-width inline chat or a floating bubble in the corner.

Read the full ChatKit deployment guide →

Channel 2: Voice AI (ElevenLabs)

Give your agent a voice. Customers call a phone number, speak to a natural-sounding AI receptionist, and book, reschedule, or cancel appointments — all by voice.

ElevenLabs handles the speech-to-text and text-to-speech. Second Brain OS connects via MCP, giving the voice agent access to the exact same tools your website chat uses. The agent speaks before executing a tool (so the caller isn’t left in silence), plays a typing sound during tool execution, and confirms the result naturally.

What you configure:

• Voice selection and conversational mode (V3 Conversational recommended for natural feel)
• System prompt with timezone injection
• SBOS MCP server connection (secret → environment variable → authorization header)
• Tool settings: post-speech execution, no-approval mode

Read the full ElevenLabs deployment guide →

Channel 3: Telegram

If your customers are on Telegram, you can deploy your agent as a Telegram bot. The setup is straightforward — it follows the standard BotFather flow, and then you register the token with Second Brain OS.

Setup:

1. Open Telegram and message @BotFather
2. Send /newbot and follow the prompts to name your bot and get a bot token
3. Go to secondbrainos.com/telegram-integration
4. Paste your bot token and click “Install Telegram Integration”

That’s it. Your Telegram bot is now connected to your Second Brain OS agent with all its tools, knowledge, and skills. Messages sent to the bot are handled by your agent and responses are sent back through Telegram.

If you’ve set up Telegram bots before with BotFather, this is the same first step — the difference is that instead of writing webhook code or connecting to a framework, Second Brain OS handles the entire backend. Your agent’s behaviour, tools, and knowledge are already configured.

Read the full Telegram deployment guide →

Channel 4: ChatGPT (Custom GPT)

If you or your customers already use ChatGPT, you can make your agent available as a Custom GPT. This is useful for internal teams who prefer ChatGPT’s interface, or for offering your agent to customers who already have a ChatGPT account.

Custom GPTs don’t support MCP servers directly. Instead, they use OpenAPI schemas to define the tools (called “Actions” in ChatGPT’s terminology). Second Brain OS generates this schema for you automatically.

Setup:

1. Go to secondbrainos.com/open-ai-custom-gpts
2. Click the button to generate your OpenAPI schema — this creates a schema containing all your public SBOS tools
3. In ChatGPT, go to Explore GPTs → Create (or edit an existing Custom GPT)
4. Under Configure → Actions, click Create new action
5. Paste the generated OpenAPI schema into the schema field
6. Under Authentication, select API Key (Bearer) and paste your Second Brain OS API key (the full Bearer recXXX:xxx-xxx-xxx value)
7. Set the model to your preferred GPT version
8. Write a system prompt that matches your agent’s behaviour — you can copy the same prompt from your SBOS profile
9. Save and publish (privately, with a link, or publicly)

If you’ve used Zapier or Make.com actions inside Custom GPTs, this is the same pattern — you’re giving ChatGPT a set of API endpoints it can call. The difference is that these endpoints are your actual SBOS tools (Calendly booking, CRM operations, knowledge search, etc.) rather than generic automation triggers.

Read the full Custom GPT deployment guide →

Channel 5: Claude Desktop & Claude Code

Claude (by Anthropic) natively supports MCP servers — which means you can connect your Second Brain OS agent directly to Claude Desktop or Claude Code with a single command. This gives Claude access to all your SBOS tools, knowledge, and skills right inside the interface you’re already using.

This isn’t just for appointment booking. Once connected, Claude can use any of your SBOS tools — search your knowledge base, create contacts, manage CRM data, run workflows, and more. It turns Claude into a fully equipped operator for your business.

Setup:

1. Go to secondbrainos.com/claude-model-context-protocol
2. Copy your user-specific npx command — it looks like this:

npx secondbrainos-mcp-server@latest rec1234 enjsefnjsnjf-efisejuf-esfkjn

3. Run the command in your terminal — it automatically installs the MCP server configuration for both Claude Desktop and Claude Code
4. Restart Claude Desktop (or start a new Claude Code session) and your SBOS tools will appear

If you’re a no-code builder using Claude to help you work, this is particularly powerful. You can ask Claude to check a client’s availability, book an appointment, search your knowledge base, or run an entire workflow — all within the conversation. No switching tabs, no manual API calls.

If you’re a developer, Claude Code with SBOS tools means you can script business operations alongside code. Deploy a function, then immediately test the booking flow it powers — without leaving the terminal.

Read the full Claude Desktop (MCP) deployment guide →

Read the full Claude Code (MCP) deployment guide →

Channel 6: CLI / Agent URL (For Developers & Custom Integrations)

The Agent URL is a raw HTTP endpoint for your Second Brain OS agent. It accepts a POST request and returns a response — no UI, no framework, just your agent over HTTP. This is the most flexible deployment option and the foundation for building custom integrations.

Your agent endpoint:

Go to secondbrainos.com/agent-url to get your user-specific curl command:

curl -X POST 'https://openai.secondbrainos.com/completion' 
  -H 'Authorization: Bearer recXXX:xxx-xxx-xxx' 
  -H 'Content-Type: application/json' 
  -d '{
    "channel": "webchat",
    "conversation_id": "conv_123",
    "session_id": "webchat_visitor_JaneDoe",
    "deliver_to_webhook": false,
    "message": "What slots are available tomorrow?"
  }'

Key parameters:

What this unlocks:

• Custom chat UIs — build your own chat interface (React, Vue, mobile app) and hit this endpoint. The channel field can be any value, so you can tag requests by source for analytics
• SaaS integrations — connect your agent to any platform that can make HTTP requests. CRM dashboards, internal tools, helpdesk systems
• Agentic workflows — use the session_id to maintain context across multiple calls. Delegate a multi-step task to your agent, then follow up in the same session. For example, a script could ask the agent to check availability, then in a follow-up message (same session) ask it to book the best slot
• Cross-platform continuity — a customer starts on WhatsApp, you pick up the same session_id in your internal dashboard, and the agent has full context of the prior conversation
• Batch operations — script bulk interactions. Process a list of leads, check availability for each, book appointments — all programmatically
• Claude Code integration — use curl within Claude Code to interact with your SBOS agent as part of a development workflow. Test your agent’s responses, simulate customer conversations, or chain agent calls with other CLI tools

If you’ve built with the OpenAI API or similar LLM endpoints, this works the same way — except the model already has your tools, knowledge, and skills attached. You don’t need to pass function definitions or manage tool calling yourself. Send a message, get a response, and the agent handles tool execution internally.

Read the full CLI & API deployment guide →

Which Channel Should You Start With?

If you’re a business owner deploying an AI receptionist:

• Start with ChatKit (website) — it’s the fastest way to get your agent live and test the booking flow visually
• Add ElevenLabs (voice) next — most customers still prefer calling, and voice has the highest conversion rate for appointment bookings
• Add Telegram if that’s where your audience is

If you’re a builder or developer:

• Start with Claude Desktop / Claude Code — test your agent’s tools interactively before deploying to customer-facing channels
• Use the CLI / Agent URL to build custom integrations and agentic workflows
• Deploy Custom GPT if your clients already use ChatGPT

The beauty of Second Brain OS is that you don’t have to choose. Configure once, deploy to all six. Each channel is independent — you can add or remove channels without affecting the others.

What’s Next

Pick a channel and follow its deployment guide:

• Website chat widget (ChatKit)
• Voice AI (ElevenLabs)
• Telegram
• ChatGPT Custom GPT
• Claude Desktop
• Claude Code
• CLI & Agent URL

All deployment guides assume you’ve completed Steps 1–4 at the top of this guide first.

This is part of a Crash Course on Second Brain OS AI Agents. Access all the Automations to deploy your AI agent to 6 channels — website, voice, Telegram, ChatGPT, Claude & CLI — you get the full masterclass when you purchase ANY AI Agent by clicking the button below

Umair Kamil

Install Second Brain OS AI Agents