Skip to content

Connect a Client

Connecting takes a minute. You point your AI client at the Dotlas MCP endpoint, authenticate, and pick your organization. After that, your assistant can use the Dotlas tools in any chat.

Before you start

Your organization must be enabled for the MCP, and you need an account that belongs to it. If you are not enabled yet, contact us to request access.

The endpoint

Point your client at this URL. It is a standard streamable-HTTP MCP server served at the bare root — there is no path to append.

https://mcp.dotlas.com

Every request is authenticated. Most clients sign in with OAuth (a browser sign-in and organization selection); headless and command-line clients can use an access key instead. Choose your method below, then jump to the recipe for your client.

Two ways to authenticate

You only ever use one of these — never both. Each recipe below lets you pick between them.

OAuth — recommended. You sign in through your browser and select an organization on the consent screen. It works in every client, including claude.ai on the web, and nothing sensitive is stored in a config file. The organization you pick scopes the data you can query, so choose the one that holds your Dotlas access.

Access key. A dl_live_ bearer key that Dotlas issues you, sent as an Authorization: Bearer <key> header. Use it for headless or command-line agents (Codex, CI jobs, scripts) or any client where a browser sign-in is awkward. A key is read-only and scoped to your organization's access, so there is no organization-selection step. Contact us to get one.

Treat an access key like a password

A key does not expire on its own. Never commit it to source control or leave it in shell history — keep it in an environment variable or your client's secret store, and contact us to rotate or revoke it if it is ever exposed.

Connection recipes

Pick your client below. Where a client supports both sign-in methods, switch between the OAuth and Access key tabs and follow just one — you do not need both.

  1. Open your connector settings in Claude and choose to add a custom connector.
  2. Paste the endpoint https://mcp.dotlas.com and confirm.

  3. When prompted, sign in through the browser window that opens.

  4. Select your organization on the consent screen. This scopes every query to that organization's licensed markets and brands.

  5. Return to Claude. The Dotlas tools now appear in the connector.

Menu labels vary by client version

Exact button and menu names change between releases. The flow is always the same: add a custom connector, use the endpoint above, sign in, and select your organization.

Using an access key on Claude Desktop

claude.ai on the web supports OAuth only. Claude Desktop can instead add a request header under the custom connector's advanced settings (beta): Authorization: Bearer dl_live_….

Opens a browser so you can sign in and select your organization:

claude mcp add --transport http dotlas https://mcp.dotlas.com

Then, inside Claude Code, run /mcp, choose dotlas, and complete the sign-in.

No browser — for headless use:

claude mcp add --transport http dotlas https://mcp.dotlas.com \
  --header "Authorization: Bearer dl_live_xxx"

Add --scope user to either command to make the server available in every project rather than just the current one.

Add the server to ~/.cursor/mcp.json (global) or .cursor/mcp.json (per project).

Cursor opens a browser to sign in on first use:

{
  "mcpServers": {
    "dotlas": {
      "url": "https://mcp.dotlas.com"
    }
  }
}

${env:DOTLAS_MCP_KEY} reads the key from your environment so it stays out of the file:

{
  "mcpServers": {
    "dotlas": {
      "url": "https://mcp.dotlas.com",
      "headers": {
        "Authorization": "Bearer ${env:DOTLAS_MCP_KEY}"
      }
    }
  }
}

In Copilot agent mode, add the server to .vscode/mcp.json.

VS Code prompts you to sign in on first use:

{
  "servers": {
    "dotlas": {
      "type": "http",
      "url": "https://mcp.dotlas.com"
    }
  }
}

Declare an input so VS Code prompts for the key once and stores it securely, then send it as a header:

{
  "inputs": [
    {
      "type": "promptString",
      "id": "dotlas-key",
      "description": "Dotlas MCP access key",
      "password": true
    }
  ],
  "servers": {
    "dotlas": {
      "type": "http",
      "url": "https://mcp.dotlas.com",
      "headers": {
        "Authorization": "Bearer ${input:dotlas-key}"
      }
    }
  }
}

Codex connects with an access key. Add the server to ~/.codex/config.toml. Reading the key from an environment variable keeps it out of the file:

[mcp_servers.dotlas]
url = "https://mcp.dotlas.com"
bearer_token_env_var = "DOTLAS_MCP_KEY"
enabled = true

Or set the header inline:

[mcp_servers.dotlas]
url = "https://mcp.dotlas.com"
http_headers = { "Authorization" = "Bearer dl_live_xxx" }
enabled = true

Run the inspector, then connect from its UI:

npx @modelcontextprotocol/inspector
  1. Set Transport Type to Streamable HTTP.
  2. Enter the URL https://mcp.dotlas.com.
  3. Complete the OAuth sign-in, or paste your access key into the Bearer Token field.
  4. Click Connect. The Dotlas tools are listed under Tools.

Any compliant MCP client connects the same way:

  1. Add a new remote / custom server with the endpoint https://mcp.dotlas.com.
  2. Authenticate — complete the OAuth sign-in in your browser (the server advertises its OAuth metadata automatically), or send an access key as an Authorization: Bearer <key> header.
  3. Select your organization when prompted (OAuth only), and the Dotlas tools appear.

Verify it works

Ask your assistant a question that uses a tool, such as "Which markets can I access?" A short list of your licensed markets confirms the connection and that your organization scope is correct.

Troubleshooting

I get a 401 or the connection is refused

The server rejects requests that are not signed in. With OAuth, complete the browser sign-in and make sure you selected an organization. With an access key, check the header is exactly Authorization: Bearer dl_live_… with no extra spaces. If authentication succeeds but the connection is still rejected, your organization may not be enabled yet — contact us to check.

No tools appear after connecting

This usually means your organization is not enabled for the MCP. Contact us and we will confirm your access.

The data looks empty or wrong

With OAuth, you may have selected the wrong organization at sign-in — disconnect, reconnect, and pick the organization that holds your Dotlas access. Results are always limited to the markets and brands your organization is licensed for.

My access key stopped working

A key is revoked the moment it is rotated. If you regenerated or reported a key, contact us for the current one, and update the header or environment variable in your client's config.

Next

Now see what you can ask, or review security and trust before you roll it out to your team.