TeleBoost Now Speaks MCP: Connect Claude, ChatGPT & Your Stack to Telegram

For as long as TeleBoost has existed, everything it knows about your Telegram pipeline (the leads you scraped, the campaigns you ran, the replies you earned) has lived inside the app. That's fine right up until the moment you want your data to do something the dashboard doesn't: sync into a warehouse, trigger an internal alert, or let an AI assistant work your inbox for you.

Today that wall comes down. TeleBoost now ships three ways to reach your pipeline programmatically: a hosted MCP server for AI tools like Claude and ChatGPT, a versioned REST API, and signed webhooks. This post explains what each one is, walks through connecting them, and shows the guardrails that keep your accounts safe while you automate.

First, what is MCP, and why should you care?

MCP stands for the Model Context Protocol. It's an open standard that lets an AI assistant talk to an outside tool through a stable, permissioned interface. Think of it as a universal adapter: instead of copy-pasting data between your CRM and a chat window, you connect the two once, and the assistant can then read and act on your live data, with the exact permissions you grant, and nothing more.

The important part for Telegram specifically: this works on top of your real accounts and CRM, not a bot. A bot API only sees the rooms a bot was invited to. TeleBoost's MCP server sits on the pipeline you actually built, so when you ask an assistant to "find my hot leads who haven't been followed up," it's looking at the same data you see in the app.

What your AI can actually do

Once connected, the assistant gets a toolbox that mirrors the app. It can search and segment contacts, read a conversation's history, update a lead's status, create and update tickets, set reminders, launch or pause DM campaigns, pull analytics, and, when you allow it, send a message from a connected account. You describe the outcome; it picks the right tools in the right order.

Here's the kind of prompt that becomes a one-liner:

Behind that sentence, the assistant calls search_contacts, then create_ticket and add_note for each match, and drafts replies with get_conversation for context: all in one turn, all scoped to exactly what you allowed.

Connecting Claude in three steps

The setup is deliberately short:

1. In TeleBoost, open Settings → Developers → Connected AI apps and copy your MCP URL: https://mcp.teleboost.app/mcp.
2. In your AI tool, add that URL as a custom MCP connector. In Claude, that's Settings → Connectors → Add custom connector.
3. Authorize. You'll land on a TeleBoost consent screen, choose which scopes to grant and whether the connection uses your solo data or a specific team, and approve. That's it: no API key to paste, no secret to store.

Authorization uses OAuth 2.1 with PKCE, so no long-lived secret ever lives on the AI tool's side, and you can revoke any connection from either end. Tools that don't speak OAuth, like n8n or a shell script, can authenticate with an API key instead.

The REST API, for everything that isn't a chat

Not every workflow is an AI conversation. The same surface is available as a clean, versioned REST API. Create a key under Settings → Developers → API keys (you'll see it once, so copy it then), and call the API with a Bearer token:

curl https://teleboost.app/api/v1/contacts?engagement_tier=hot \
  -H "Authorization: Bearer tb_live_your_key_here"

# → { "data": [ { "id": 1842, "username": "maria_k",
#       "lifecycle_status": "replied", "engagement_tier": "hot" }, ... ],
#     "next_cursor": "eyJpZCI6MTg0Mn0", "has_more": true }

Every list endpoint uses cursor pagination and returns the same envelope, so paging through thousands of contacts is a loop on next_cursor. The full schema is published as OpenAPI at https://teleboost.app/api/v1/openapi.json, and the developer docs cover every resource with examples.

Sending a message is one call, and it runs through the exact same pipeline as the app:

curl -X POST https://teleboost.app/api/v1/messages \
  -H "Authorization: Bearer tb_live_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{ "account_id": 3, "contact_id": 1842,
        "text": "Hey Maria — following up on the demo you asked about." }'

Webhooks: stop polling, get pushed

Polling an API to notice when a lead replies is wasteful. Webhooks flip it around: register an HTTPS endpoint under Settings → Developers → Webhooks, pick the events you care about, and TeleBoost calls you the moment they happen: a reply lands, a ticket is created, a campaign finishes.

Every delivery is signed so you can trust it. Verify the X-TeleBoost-Signatureheader, which is an HMAC-SHA256 of the request body using your endpoint's secret:

import hmac, hashlib

def verify(request_body: bytes, signature_header: str, secret: str) -> bool:
    expected = "sha256=" + hmac.new(
        secret.encode(), request_body, hashlib.sha256
    ).hexdigest()
    return hmac.compare_digest(expected, signature_header)

Deliveries retry with exponential backoff for up to 24 hours, and an endpoint that keeps failing is auto-disabled with an alert, so a broken receiver never silently drops events or hammers your server. Route them into n8n, Make, Zapier, or your own backend.

Access you control, down to the scope

Power without control is a liability, so every credential is least-privilege by default. When you create a key or authorize an app, you choose exactly which scopes it gets, from read-only to CRM automation to full access, and whether it works on your solo data or a specific team. Keys and tokens are hashed at rest, shown only once, and revocable instantly. Session strings and proxy credentials are never exposed, not even to admins. And every request is rate-limited per plan and written to an audit log you can review.

Getting started

The API, MCP server and webhooks are available on the Pro and Business plans. Head to Settings → Developers, create your first key or connect an AI app, and skim the developer docs for the full reference. If you're weighing whether this fits your workflow, the MCP & API overview walks through the capabilities and the security model in plain terms.

Your pipeline was never meant to be trapped behind a screen. Now it's one connection away from everything else you use.