Skip to content
English
  • There are no suggestions because the search field is empty.

How Do I Connect Dataslayer MCP to Any MCP-Compatible LLM (Custom)?

Connect Dataslayer MCP to any MCP-compatible client or LLM, including local, self-hosted models. Universal setup: server URL, OAuth 2.1, and a remote MCP connection.

Dataslayer MCP isn't limited to the named providers (Copilot, ChatGPT, Claude, Mistral). Because MCP is an open protocol, you can connect Dataslayer to any MCP-compatible client or model, including local, self-hosted models running in your own environment. This "Custom" option gives you full flexibility to query your marketing data from whatever AI tooling you use.

What You'll Need Before Starting

  • An MCP-compatible client or LLM that supports remote MCP servers and OAuth authentication. Examples include desktop AI apps and IDE assistants such as Claude Desktop, Cursor, or VS Code; automation tools like n8n; and your own code built with an MCP client SDK.
  • An active Dataslayer account with at least one data source connected.
  • (For local / self-hosted models) a client or runtime that can reach the internet to complete the OAuth flow with Dataslayer.

Universal Connection Details

These are the same for every client:

  • Server URL: https://agents.dataslayer.ai/mcp
  • Authentication: OAuth 2.1
  • Transport: remote MCP over HTTP (streamable)

Step-by-Step Instructions

1. Open your client's MCP / connector settings

Go to the place where your client adds external MCP servers or connectors. The exact name and location vary by client (e.g. "Connectors", "MCP Servers", "Tools").

2. Add a new remote MCP server

  • Paste the Server URL exactly: https://agents.dataslayer.ai/mcp
  • Give it a name like "Dataslayer".

3. Choose OAuth as the authentication method

Select OAuth (Dataslayer uses OAuth 2.1).

4. Authorize with Dataslayer

Complete the OAuth login with your Dataslayer credentials. Your marketing platform credentials never leave Dataslayer.

5. Use the tools

Once connected, your client detects the Dataslayer tools and you can query your data in natural language.

 

Tip (optional): test with "Using Dataslayer, pull my Google Ads spend for the last 7 days."

Notes by Client Type

  • Desktop AI apps / IDE assistants (e.g. Claude Desktop, Cursor, VS Code): add Dataslayer as a custom/remote MCP server in the app's MCP or connectors settings.
  • Automation / workflow tools (e.g. n8n): add an MCP client node or credential pointing to the server URL with OAuth.
  • Custom code: use an MCP client SDK, point it at the server URL, and implement the OAuth 2.1 flow.
  • Local / self-hosted models: run them through any MCP-capable client. The model stays in your environment; the only thing leaving your perimeter is the encrypted MCP traffic to Dataslayer. This gives maximum data residency.

Best Practices

Be specific about what you want:

  • Mention which metrics and dimensions you need
  • Include date ranges and any filters
  • Specify the data source if you have multiple accounts connected

Example of a well-structured prompt:

Using Dataslayer, retrieve my Google Ads data for account "{YOUR-ACCOUNT-NAME}" for the last 30 days. Show spend, impressions, and clicks grouped by campaign name, sorted by spend from highest to lowest.

Troubleshooting Common Issues

  • Your client doesn't support the connection: confirm it supports remote MCP servers and OAuth (not only local stdio servers without authentication).
  • Connection fails: check the Server URL is exactly https://agents.dataslayer.ai/mcp, with no extra spaces.
  • Authentication doesn't complete: verify your Dataslayer credentials; your client must be able to open the OAuth browser flow to authorize.
  • No tools appear: disconnect and reconnect, and make sure the connection is enabled/authorized in your client.

Authentication and Security

Every client connects to Dataslayer through a secure, encrypted connection using the Model Context Protocol and OAuth 2.1. Your Dataslayer login and marketing platform credentials are never shared with the model or client — all queries are processed through Dataslayer's secure infrastructure. With a local / self-hosted model, data residency is maximal.

For more details, see How does Dataslayer MCP handle authentication? and Is my data secure when using MCP with external AI models?

Usage Limits

Dataslayer's shared limit of 50 API calls per day applies regardless of which client you use. Any message or rate limits beyond that depend on the client/model you choose; with a local / self-hosted model they're governed by your own infrastructure. See Is there a limit to the number of requests per AI provider?

Comparing AI Providers

Custom is one of several ways to use Dataslayer MCP. To compare with Copilot, ChatGPT, Claude, and Mistral:

 

As always, please contact us via our live chat on our website or via email if you still have doubts or questions. We are happy to help!