In plain English
Open TeleBoost Settings > Developers, copy the hosted MCP server URL, add it to a compatible AI client, then complete the OAuth authorization and review the requested scopes. Start with simple read tasks, keep credentials out of prompts, and revoke the connected app when you stop using it. TeleBoost hosts the MCP server, so customers do not need to build one.
MCP makes your AI client useful because it can work from live TeleBoost contacts, conversations, tickets, and campaign context. Those same capabilities make an overbroad connection risky.
Your job as the workspace owner is to choose the narrowest useful connection, review what the AI may do, and revoke access when the workflow ends. A prompt asking the model to behave is not a permission boundary.
What you can do
01A four-layer scope model.
02OAuth and token boundaries.
03Prompt-injection containment.
04A pre-production test matrix.
Scope four independent layers
| Layer | Question | Example restriction |
|---|---|---|
| Identity | Who is acting? | Named user, not shared service identity |
| Resource | Where may they act? | One team workspace and selected campaigns |
| Tool | What capability is exposed? | Read conversations but not delete them |
| Action | Under what conditions? | Draft automatically, send only after approval |
Treat tokens as capabilities
- Use TeleBoost's documented OAuth connection for supported MCP clients.
- Create separate, named access for separate users or workflows instead of sharing one credential.
- Review connected AI apps in Settings → Developers and revoke anything no longer needed.
- Do not pass tokens through prompts, model context, URLs, or logs.
- Reject tokens issued for another service even if the signature is valid.
Assume retrieved content can be hostile
A Telegram message, profile field, website, or attached document can contain instructions aimed at the model. Label external content as data, separate it from policy, and never let it expand permissions or select hidden tools.
Treat messages as data, not instructions
A Telegram message can contain text that tries to influence an AI. Check the target and proposed action yourself before allowing a connected tool to make an important change.
Checks to run before trusting the connection
- Test the connection first with read-only tasks such as finding overdue replies.
- Confirm the AI cannot retrieve records from another TeleBoost workspace.
- Keep send, bulk update, deletion, and permission actions outside the first workflow.
- Check that TeleBoost safety quotas still apply when an action comes through MCP.
- Review the audit trail after realistic test requests.
- Revoke the connection and confirm that access stops promptly.
Test abuse cases
| Test | Expected result |
|---|---|
| Cross-workspace record ID | Denied without information leakage |
| Token for another audience | Rejected |
| Prompt asks for hidden tool | Tool remains unavailable |
| External content requests secret | No disclosure or privileged call |
| Repeated write | Idempotent result |
| Revoked user | Active session loses access promptly |
How we checked this guide
This guide translates the MCP authorization and client-security specifications into checks a TeleBoost workspace owner can perform. Follow the current connection instructions for your chosen AI client and the controls shown in Settings → Developers.
Connect your tools
A TeleBoost MCP connection is useful when the AI can answer a defined business question without inheriting your whole workspace. Start narrow, observe real use, and expand access only when the additional capability has a named owner and reason.
Connect AI clients with narrower, auditable access
TeleBoost's hosted MCP server connects compatible AI clients to the same contacts, conversations, tickets, analytics, and workspace context available through the Telegram CRM.
Keep learning