Skip to main content

Connect AI SmartTalk to any MCP client

AI SmartTalk exposes a hosted MCP server, so you can manage your assistants from any AI tool that speaks the Model Context Protocol (MCP) — not just Claude. Claude Desktop, Claude on the web, Claude Code, ChatGPT, Cursor, VS Code (GitHub Copilot), Cline, Windsurf, and any custom MCP client all work the same way.

You stay inside the AI tool you already use and ask it, in plain language, to list assistants, read or rewrite their instructions, search and grow their knowledge base, and diagnose a client's setup. There's nothing to install and no API key to copy — you sign in once with your usual AI SmartTalk account over a secure OAuth flow.

This is the opposite direction of the MCP Servers connector. There, your assistant calls external tools. Here, an external AI tool calls your AI SmartTalk assistants.


How it works​

The MCP client connects to a single secure endpoint, you approve access with your AI SmartTalk login, and the AI tool can then call AI SmartTalk tools on your behalf. Every action is scoped to your account — you only ever see and act on the assistants you're already allowed to manage in your dashboard.


The single endpoint​

Every MCP client connects to the same address:

https://aismarttalk.tech/api/mcp
  • Transport: Streamable HTTP (JSON-RPC 2.0)
  • Authentication: OAuth 2.0 — you sign in with your AI SmartTalk account; no API keys to manage
  • Security: HTTPS only; access limited to the assistants your account manages; revocable at any time

What you can do​

Once connected, your AI tool can use these AI SmartTalk capabilities automatically, picking the right one from what you ask:

ActionWhat it does
List assistantsList every assistant your account can manage
Find assistantsLocate an assistant by name, by a user's email, or by organization
Read instructionsRead an assistant's biography — its personality and decision rules
Update instructionsRewrite an assistant's instructions and routing logic
Search knowledgeRun a natural-language search across an assistant's knowledge base
Browse knowledgeExplore the knowledge base interactively in a built-in viewer
Add a documentDrop in a new knowledge document, indexed automatically
Add FAQ entriesAdd one or several question/answer pairs at once
Run diagnosticsInspect an assistant's plan, AI model, knowledge size and connected channels
List channels & integrationsSee which channels and connectors an assistant is wired to
Rename an assistantChange an assistant's display name

It also ships ready-to-use prompts your client can surface directly — for example "improve this assistant's biography" and "audit knowledge coverage for a topic".

You can always refer to an assistant by its name — no need to look up an ID.


Use cases​

Support teams — fix a client's assistant without leaving your AI tool​

"A client with the email [email protected] reported their bot gives wrong refund answers. Find their assistant, check how it's set up, and search its knowledge base for the refund policy."

The AI tool finds the assistant by email, runs diagnostics (plan, model, channels, knowledge size), and surfaces the relevant knowledge — so you can triage a ticket in seconds.

Agencies — manage many client assistants from one place​

"List all assistants in the Acme organization and tell me which ones have fewer than 10 knowledge documents."

Perfect when you manage dozens of assistants across clients: enumerate them, spot the ones that need attention, and act — all from Claude or Cursor.

Content & knowledge ops — keep the knowledge base fresh​

"Add an FAQ to my Support assistant: 'How long do refunds take?' — 'Refunds arrive within 5 business days.'" "Audit my Sales assistant's knowledge coverage for 'enterprise pricing' and tell me what's missing."

Grow and audit the knowledge base conversationally, and let the AI tool draft the documents or FAQ entries for you before indexing them.

Prompt engineering — refine instructions in a loop​

"Read my Support assistant's instructions, then rewrite them to answer in a friendly, concise tone and always escalate billing questions to a human."

Iterate on an assistant's personality and routing rules with the AI tool doing the heavy lifting, and see the changes appear in your AI SmartTalk dashboard instantly.

Developers — script assistant management from your IDE​

From Cursor or VS Code: "Bootstrap a new FAQ set for the Onboarding assistant from this README."

Manage assistants right next to your code, without switching to the dashboard or wiring up the REST API.


Before you start​

  • An active AI SmartTalk account
  • An AI tool that supports MCP connectors (see the list below)

No API key, no technical setup — you sign in with your usual AI SmartTalk account.


Connect your AI tool​

Claude (Desktop or web)​

  1. Open Claude's Settings → Connectors.

  2. Choose Add custom connector.

  3. Enter the AI SmartTalk MCP address:

    https://aismarttalk.tech/api/mcp
  4. Confirm. Claude opens an AI SmartTalk sign-in page.

  5. Sign in and approve access. Claude is now connected to your assistants.

Claude Code​

In your terminal:

claude mcp add --transport http aismarttalk https://aismarttalk.tech/api/mcp

Then start Claude Code, run /mcp, and follow the sign-in link in your browser to approve access.

Cursor​

  1. Open Cursor Settings → MCP → Add new MCP server.

  2. Add the AI SmartTalk server:

    {
    "mcpServers": {
    "aismarttalk": {
    "url": "https://aismarttalk.tech/api/mcp"
    }
    }
    }
  3. Cursor opens the AI SmartTalk sign-in page — approve access.

VS Code (GitHub Copilot)​

  1. Run MCP: Add Server from the Command Palette, or add the server to your .vscode/mcp.json:

    {
    "servers": {
    "aismarttalk": {
    "type": "http",
    "url": "https://aismarttalk.tech/api/mcp"
    }
    }
    }
  2. Start the server from the MCP view and complete the sign-in in your browser.

Cline, Windsurf & other clients​

Any client that supports a remote (HTTP) MCP server with OAuth works the same way. Add a server with:

  • Name: aismarttalk
  • Transport / type: http (Streamable HTTP)
  • URL: https://aismarttalk.tech/api/mcp

When the client first calls the server, it will open the AI SmartTalk sign-in page so you can approve access.

Custom / your own MCP client​

The server implements the standard MCP Streamable HTTP transport and acts as an OAuth 2.0 resource server. Point your client at https://aismarttalk.tech/api/mcp; an unauthenticated request returns 401 with a WWW-Authenticate header pointing at our protected-resource metadata, so a compliant MCP client can run the OAuth flow automatically.


Try it​

Ask your AI tool things like:

  • "List my AI SmartTalk assistants."
  • "Show me the instructions for my support assistant."
  • "Update my assistant so it always answers in a friendly, concise tone."
  • "Search my Support assistant's knowledge base for the refund policy."
  • "Add an FAQ to my Support assistant: 'How long do refunds take?' — 'Refunds arrive within 5 business days.'"
  • "A client with the email [email protected] has an issue — find their assistant and tell me how it's set up."

Security​

  • Access is granted through your own secure AI SmartTalk OAuth sign-in — there are no keys to copy or store.
  • The connection only reaches the assistants your account already manages.
  • The endpoint works over HTTPS only.
  • You can revoke access at any time from your AI tool's connector settings.

FAQ​

Does this only work with Claude? No. Any AI tool with an MCP client works — Claude, ChatGPT, Cursor, VS Code (Copilot), Cline, Windsurf, and custom clients. Claude is just the most common example.

Do I need an API key? No. You authenticate with your normal AI SmartTalk account through OAuth. There are no keys to manage.

Which assistants can the AI tool see? Only the ones your account is already allowed to manage — exactly as in your AI SmartTalk dashboard.

Are changes live? Yes. Edits you make through your AI tool — like updating instructions or adding knowledge — appear in your AI SmartTalk dashboard right away, and changes in the dashboard are visible to the AI tool too.

What's the difference with the MCP Servers connector? The MCP Servers connector lets your assistant call external tools. This page is the reverse: an external AI tool calls your AI SmartTalk assistants.


Good to know​

  • The connection works on a secure (HTTPS) address only.
  • Changes you make through your AI tool appear in your AI SmartTalk dashboard right away, and the other way around.
  • You can connect the same AI SmartTalk account from several AI tools at once.

Ready to elevate your
user experience?

Deploy AI assistants that delight customers and scale with your business.

GDPR Compliant